How big is AI compared with a brain? 1,092 models against 11 animals

01 Key findings

1. The largest model with a published size is cat-sized, not human-sized. 3 trillion parameters against 2.2 to 2.4 trillion synapses in a cat neocortex and 150 to 164 trillion in a human one: a gap of about 1.7 orders of magnitude, 50 to 55 times.Epoch AI; IQ Metrics calculation
2. AI passed a roundworm in 1975 and a fruit-fly larva in 1980. The first notable model above a mouse neocortex that is a real neural network appeared in 2019; a cat followed in 2025.IQ Metrics calculation
3. Half of all notable models are at least as big as our low estimate for a honeybee (379 of 728), 128 are bigger than a mouse neocortex, 6 reach a cat, none reach a macaque or a human.IQ Metrics calculation
4. Growth has stalled since 2021. The largest model doubled every 1.2 years from 2012 to 2026, but 1.6 trillion (2021) became only 3 trillion (2025).Epoch AI; IQ Metrics calculation
5. Parameters are not synapses. Most trillion-parameter models are sparse: they use 3% to 16% of their parameters at once, and one real cortical neuron needed a five- to eight-layer network to copy.Epoch AI; Beniaguev et al. 2021
6. The biggest models are the least documented. A size is printed for 41% of the models of 2026, against 92% in 2021.IQ Metrics calculation

02 Statistics to quote

03 How big is AI compared with a human brain?

The largest AI model with a published size has about 3 trillion parameters, roughly 50 to 55 times fewer than the 150 trillion to 164 trillion synapses counted in a human neocortex, and level with a cat’s neocortex (2.2 trillion to 2.4 trillion synapses by our calculation) while just under a macaque’s (3.1 trillion to 6.5 trillion) ^{[1] [15] [14]}. This page sets all 1,092 notable models in Epoch AI’s table against the synapse counts of 11 animals, so that a number such as “3 trillion” can be read on the same scale as a brain.

The comparison is about size only. A parameter is one number a model learns; a synapse is a physical contact between two neurons. Nothing here says that any model thinks like a cat, a mouse or a person, and the scale of a brain is not its ability: a crow, a honeybee and a human with half a brain all do more than their synapse counts suggest. Our own dataset joins two sources that are rarely put side by side: Epoch AI’s table of notable models (CC BY 4.0), and synapse counts from primary papers, from electron-microscopy reconstructions of whole fly and worm brains to the stereology of a human cortex. Where papers disagree we keep the range and show it ^[1].

Figure 1

Every notable AI model against the synapses of eight animals: the biggest models have just reached a cat

Each blue dot is one of 728 notable AI models with a parameter count (Epoch AI, 1950 to 2026): publication date across, size up as a power of ten (6 = a million, 12 = a trillion). Orange dots at the right are animal brains, plotted at the middle of their synapse range; the largest model, 3 trillion, sits level with a cat and just under a macaque, 50 times under a human neocortex. As of 10 November 2026.

AI model (parameters)Sparse or mixture-of-experts model (total parameters)Count of n-grams or features, not network weights

Size: parameters (models) or synapses (animals), as a power of ten

03691215

RoundwormFly larvaFruit flyHoneybeeMouseCatMacaqueHuman

195019802010

Year published (animals at the right have no date)

Animals: neocortex for the mammals, whole brain or body for the small animals; one parameter is not one synapse.

Three reference points show how the page reads. The Perceptron Mark I, built at the Cornell Aeronautical Laboratory in 1957, had about 1,000 adjustable weights ^[1]; AlexNet, a 2012 image network, had 60 million; BERT-Large (2018) had 340 million; GPT-3 (2020) had 174.6 billion; and Switch Transformer (2021) had 1.57 trillion. On the same scale a roundworm has 5,000 to 8,000 synapses, an adult fruit fly 46 to 312 million, a mouse neocortex 81 billion and a human neocortex 150 to 164 trillion. Our guide to human brain facts gives the other numbers people ask about, and human brain vs dog brain compares neuron counts for two familiar species.

Parameter

One learned number in an AI model, a weight or a bias. Epoch AI records the “number of learnable parameters in the model” ^[2].

Synapse

A contact where one neuron signals to another. Counting rules differ: some papers count presynaptic sites, some postsynaptic densities, some connections between pairs of neurons.

Neocortex

The folded outer layer of the mammalian brain. We use it for mammals because it is the part for which synapse totals were counted.

Connectome

A complete map of the neurons of a nervous system and the synapses between them, reconstructed from electron-microscope images ^{[3] [9]}.

Mixture of experts (MoE)

A model built from many sub-networks of which only a few run for each token. Its total parameter count is large; its active count is much smaller ^[2].

Active parameters

The parameters used to process one token in a sparse model. The Epoch AI column holds the total, with the active number in the note when it is known.

Order of magnitude

A factor of ten. The gap between the largest model and a human neocortex is 1.7 orders of magnitude, a factor of about 50.

Notable model

A model in Epoch AI’s table that is highly cited, expensive to train, widely used, state of the art or historically important ^[2].

04 Is an AI parameter the same as a brain synapse?

No: a parameter is a single stored number, while a detailed model of one cortical neuron needed a deep network to copy. A 2021 study trained artificial networks to copy a biophysical model of one layer 5 pyramidal neuron and found that a network with five to eight layers was needed to capture its input and output ^[24]. So counting one synapse as one parameter probably understates the brain; counting a trillion parameters of a sparse model as a trillion working connections probably overstates the model.

Several differences make the match loose. A synapse changes with time and activity; a trained model’s weights are fixed when it is used. A synapse is counted in different ways (a presynaptic site that talks to six cells is one synapse in one paper and six connections in another). Brains also differ from models in what else they carry: neurons, glia, wiring length, energy use, a body and a lifetime of learning. We use parameters and synapses because both are counts of adjustable links, and because they are the only numbers published on both sides. For what AI scores on tests built for people, see the ChatGPT IQ score and why a model’s “120” is not an IQ; our study AI against the maths olympiad tests ability directly.

05 How many neurons and synapses does a human brain have?

A human brain has about 86.1 billion neurons, 19% of them (about 16 billion) in the cerebral cortex, and the neocortex alone holds 150 to 164 trillion synapses; nobody has counted the synapses of the whole brain ^{[18] [15] [14]}. The neuron figure comes from the isotropic fractionator, which dissolves brain tissue and counts cell nuclei, applied to four adult men; it replaced the textbook figure of 100 billion, which the authors say was never a count; the synapse figures come from counting synapses in electron-microscope samples of five young men and multiplying by the volume of the neocortex, a measurement with a coefficient of variation of 0.17 between brains ^[14].

The two human synapse figures, 164 trillion in 2001 and 0.15 quadrillion in 2003, are not two independent checks: the 2003 figure appears in a review of neurostereology studies, and we found no second, independent counting study, so we treat them as one estimate of 150 to 164 trillion. Round figures such as “100 trillion synapses” and “1,000 trillion” appear in textbooks and articles; we found no counting study behind either. The cerebellum, the thalamus and the rest of the brain are not counted at all, so a human brain holds more synapses than the neocortex figure, and every comparison with a mammal on this page is a lower bound for the whole brain. A single cubic millimetre of human temporal cortex, reconstructed in 2024, holds 133.7 million synaptic connections and 57,216 cells ^[12]; our study of how much of the brain has been mapped shows how little of any brain is reconstructed so far, and how much data the brain can store covers the memory question.

Figure 2

Synapses from a roundworm to a human neocortex: 14 orders of magnitude; one fruit fly has counts 7 times apart, our honeybee bracket 21 times

Low to high published or calculated synapse count for 11 animals, as a power of ten (3 = a thousand, 9 = a billion, 12 = a trillion). Bar = low to high; dot = the headline count where a paper prints one (fruit fly: 54.5 million), otherwise the middle of the range. Mammals are neocortex only. The honeybee, cat and macaque are our calculations, not counts. As of 10 November 2026.

Sea squirt larva: 4,846 to 6,6183.8 (3.7–3.8)

Roundworm: 5,000 to 8,0003.8 (3.7–3.9)

Marine worm larva: 26,881 to 32,3814.5 (4.4–4.5)

Fruit fly larva: 544,0005.7 (5.7–5.7)

Zebrafish larva: 25.25 million7.4 (7.4–7.4)

Fruit fly: 46 million to 312 million7.7 (7.7–8.5)

Honeybee: 391 million to 8.2 billion9.3 (8.6–9.9)

Mouse cortex: 81 billion10.9 (10.9–10.9)

Cat cortex: 2.2 trillion to 2.4 trillion12.4 (12.3–12.4)

Macaque cortex: 3.1 trillion to 6.5 trillion12.7 (12.5–12.8)

Human cortex: 150 trillion to 164 trillion14.2 (14.2–14.2)

3691215

Every animal on the ladder has its own caveat, listed in the table at the end of the page. The roundworm and the fruit-fly larva are counted by electron microscopy, synapse by synapse, and are the most exact rows; the mouse and human rows are densities times volumes; the cat and macaque rows are our calculation from neuron counts and the number of synapses per neuron measured in one cortical area; and the honeybee is a bracket, because we found no paper that prints a synapse count for a honeybee brain.

06 How many synapses do a fruit fly, a honeybee and a mouse have?

An adult fruit fly brain has 54.5 million synapses between 139,255 neurons, a honeybee brain has no published count but about one million neurons, and a mouse neocortex has about 81 billion synapses ^{[9] [21] [13]}. The FlyWire consortium reconstructed the fruit fly brain from electron-microscope images and counted 5 x 10⁷ chemical synapses; a separate reconstruction of the whole male fly central nervous system detected 46 million presynaptic sites connected to 312 million postsynaptic densities ^[10]. Same animal, three counts from 46 million to 312 million, largely because a synapse with several partners can be counted at its source or at each target. We show the whole range.

The honeybee is the row readers ask about most, and the one with the weakest data. Reviews say its brain has “approximately one million neurons” or “less than a million”, and its mushroom bodies hold about 368,000 Kenyon cells, more than 40% of the brain’s neurons ^{[21] [22] [23]}. No paper we found prints a total for its synapses, so we bracket it: one million neurons times the 391 synapses per neuron measured in the fly brain gives 391 million, and one million neurons times the 8,200 synapses per neuron of the mouse neocortex gives 8.2 billion. That bracket, 391 million to 8.2 billion, is a calculation by IQ Metrics, not a count. The one billion often quoted for the bee in blog posts sits inside it; we found no counting study behind that figure. See also what bees, ants and flies can do.

Smaller animals are better counted. The roundworm Caenorhabditis elegans has 302 neurons and about 5,000 chemical synapses in the original reconstruction, and about 8,000 in adults in a later one ^{[3] [4]}. A sea-squirt tadpole has 177 neurons and 6,618 synapses, 1,772 of them onto muscle ^[5]; a three-day marine worm larva has 26,881 synapses in its published connectome ^[6]; a first-instar fruit-fly larva has 3,013 neurons and 544,000 synapses ^[7]; and a six-day zebrafish larva has 25.25 million synapses detected automatically in a preprint ^[8]. One cubic millimetre of mouse visual cortex holds 524 million synaptic clefts, more than a language model of 2018 had parameters ^[11].

07 When did an AI model get as big as a fruit fly, a mouse or a cat?

Notable AI models matched a roundworm’s synapses in 1975, a fruit-fly larva’s in 1980, an adult fruit fly’s between 1995 and 2012 depending on which count and which kind of model is used, a mouse neocortex’s in 2019 to 2020, and a cat neocortex’s in 2025 ^[1]. The first row of Epoch’s table above each animal’s low synapse count is the earliest date; we also give the first real neural network, and the first year with three or more such models, so that one odd row cannot set the year.

The odd rows are worth naming. The first model above a roundworm is the Cognitron, a Japanese network of 1975 with 21,600 learnable weights; the first above a fruit-fly larva is the Neocognitron of 1980, from NHK’s research laboratories in Japan, with 1.14 million ^[1]. But the first row above a fruit fly (1995) is a support-vector machine whose 100 million parameters are, in Epoch’s note, the dimension of a polynomial feature space, and the first rows above a honeybee and a mouse (2007) are two Google n-gram tables, SB-LM with 300 billion entries and KN-LM with 21 billion, which count word sequences and are not neural networks. We keep them in the table, mark them, and show the first neural network too: 2009 for a fruit fly, 2012 for a honeybee, and 2019 for a mouse, the recommender DLRM at 100 billion, with GPT-3 at 174.6 billion following in 2020 ^[1].

Figure 3

The year AI got as big as each animal: a roundworm in 1975, a cat in 2025, nothing bigger yet

For each animal, the first notable AI model whose parameter count is at or above the low end of its synapse range (orange), the first real neural network if an n-gram or feature count came earlier (blue), and the first year with three or more such models (navy). Mouse: the 2007 n-gram table SB-LM comes first, GPT-3-era networks in 2019 and 2020. Macaque and human: no model yet. As of 10 November 2026.

First notable model at or above the animalFirst neural network at or above itFirst year with 3 or more models

Sea squirt larva4,846 to 6,618 synapsesCognitron 19753+ models in 1994

Roundworm5,000 to 8,000 synapsesCognitron 19753+ models in 1994

Marine worm larva26,881 to 32,381 synapsesNeocognitron 1980

Fruit fly larva544,000 synapsesNeocognitron 19803+ models in 2009

Zebrafish larva25.25 million synapsesSupport Vector Machines 1995

Fruit fly46 million to 312 million synapsesSupport Vector Machines 1995

Honeybee391 million to 8.2 billion synapsesKN-LM 2007

Mouse cortex81 billion synapsesSB-LM 2007

Cat cortex2.2 trillion to 2.4 trillion synapsesGrok 3 2025

Macaque cortex3.1 trillion to 6.5 trillion synapses – not reached

Human cortex150 trillion to 164 trillion synapses – not reached

1960197019801990200020102020

The cat is the newest rung. Grok 3, published in February 2025 with 3 trillion parameters (Epoch confidence “Likely”), is the first notable model at or above the low estimate for a cat neocortex, 2.2 trillion; Grok 4 ties it. 6 models in total reach that size and 4 reach the cat’s high estimate of 2.4 trillion. No model reaches the low estimate for a macaque neocortex (3.1 trillion) – the largest is 4% short – or for a human. Our page on cat IQ covers what is known about how cats think; this page says nothing about it, only that the numbers of the two have the same order of magnitude.

08 How many AI models are bigger than a honeybee or a mouse?

Of the 728 notable models with a parameter count, 379 (52%) are at or above the low estimate for a honeybee, 128 (18%) at or above a mouse neocortex, 6 at or above a cat and none at or above a macaque or a human ^[1]. Another 364 notable models have no parameter count at all, so these shares are shares of the models that print one; the unprinted models include most recent frontier systems.

Figure 4

From 1,092 models to 128 above a mouse, 6 above a cat and none above a human

All 1,092 notable models in Epoch AI's table, the 728 with a parameter count (the other 364 print none), and how many of those are at or above the synapse count of five animals. The five rungs use the low estimates; no model reaches a macaque (3.1 trillion) or a human. As of 10 November 2026.

Notable AI models in the Epoch AI table1,092all

364 notable models have no parameter count (most of them from labs that do not publish one)

With a published or estimated parameter count72866.7% of all

Size of the adult fruit fly brain, 54.5 million synapses

At or above the synapses of a fruit fly (54.5 million)51046.7% of all

Honeybee: our calculation, no count exists

At or above a honeybee (391 million, low estimate)37934.7% of all

Mouse neocortex, counted by Schuz and Palm

At or above a mouse neocortex (81 billion)12811.7% of all

Cat neocortex, our calculation

At or above a cat neocortex (2.2 trillion, low estimate)60.5% of all

Human neocortex, 150 trillion counted

At or above a human neocortex (150 trillion, low estimate)00.0% of all

The typical notable model is small. The median has 558 million parameters, inside the honeybee range of 391 million to 8.2 billion, and 335 models have a billion parameters or more; 127 have 100 billion or more, and 27 reach a trillion ^[1]. Apart from two 2007 n-gram tables, models of at least a billion parameters first appear in 2012 (a Google image network with a billion connections) and make up 73% of the notable models of 2020 to 2026 that print a size.

Figure 5

Most notable AI models are small: the middle one has 558 million parameters, inside the honeybee range

How the 728 notable models with a parameter count spread across sizes, in bins of one power of ten (7 = ten million, 11 = a hundred billion). Median 558 million; 335 models have at least a billion parameters and 27 at least a trillion. Dotted lines: the median model, the adult fruit fly and the mouse neocortex. As of 10 November 2026.

6102322235111814010610210027Median modelFruit flyMouse

135791113

Parameters, as a power of ten

By period, the median model grew from about 4,100 in the 1980s to 55 billion in 2023-2026, but the spread widened: a notable model of 2023 to 2026 can have 57,000 or 3 trillion parameters. Before 2020 only 2 models reached the mouse tier (the 2007 n-gram table and the 2019 recommender DLRM), and from 2023 to 2026 101 of the 244 models with a count did ^[1].

Figure 6

Size by era: the typical notable model grew from thousands of parameters to tens of billions

One dot per model (728), grouped by publication period; box = middle half, whiskers = smallest and largest, sizes as a power of ten. The median rose from about 4,100 in the 1980s to 55 billion in 2023 to 2026; the dotted line is the mouse neocortex. As of 10 November 2026.

Middle halfMedianWhiskerslog10 parameters

Before 198012 models2.8

1980s15 models3.6

1990s29 models4.1

2000s24 models6.6

2010-201455 models7.6

2015-2019161 models7.7

2020-2022188 models9.0

2023-2026244 models10.7

135791113

Parameters, as a power of ten. Dotted line: Mouse neocortex.

Figure 7

Size tier by period: before 2020, 88% of models sat below a honeybee; since 2023, 101 of 244 sit above a mouse

Number of notable models per size tier (the animal they have reached) and publication period. Tiers are cut at the fruit fly, honeybee, mouse neocortex and cat neocortex. In 2023 to 2026, 101 models are above a mouse and 124 are in the honeybee-to-mouse tier. Each column starts at the animal named: the first holds models below a fruit fly, "fly" is fruit fly to honeybee, "bee" honeybee to mouse, "mouse" mouse to cat and "cat+" cat or more. As of 10 November 2026.

0 models124 models

<fly<flyflyflybeebeemousemousecat+cat+Before 19801200001980s1500001990s2810002000s2021102010-201432167002015-2019835225102020-20222148942502023-2026712124956

Figure 8

Notable models per year by size tier: models above a mouse arrive from 2019, and 2023 was the biggest year

Notable models with a parameter count per publication year, 2012 to 2026, split by which animal they have reached. Counts rise to 91 in 2023 and fall after, partly because fewer labs print a size (see the section on missing sizes). 2026 runs to 28 October. As of 10 November 2026.

Below a fruit flyFruit fly to honeybeeHoneybee to mouseMouse to catCat or more

9'12

12'13

21'14

20'15

30'16

25'17

30'18

56'19

43'20

73'21

72'22

91'23

56'24

58'25

39'26

09 How fast is the largest AI model growing, and when would it match a brain?

The largest model published each year doubled about every 1.2 years from 2012 to 2026, but from 2021 to 2026 it barely moved: 1.57 trillion (Switch, 2021) became 3 trillion (Grok 3, 2025) ^[1]. A straight line through the 15 yearly maxima rises x1.76 a year (R squared 0.72); using only rows Epoch rates Confident or Likely gives a doubling every 0.98 years. The window matters: 2018 to 2025 gives 0.99 years, 2021 to 2026 gives 3.4 years with an R squared of only 0.27.

Figure 9

The largest model doubled about every 1.2 years from 2012 to 2026, but has barely grown since 2021

The largest notable model with a parameter count in each year, 2012 to 2026 (size as a power of ten), with the straight least-squares line through all 15 points: x1.76 a year, a doubling every 1.23 years (R squared 0.72). With only Confident and Likely rows the doubling is 0.98 years; from 2021 to 2026 alone it is 3.4 years (R squared 0.27). 2026 runs to 28 October. As of 10 November 2026.

Largest model, parameters as a power of ten

8910111213

DistBelief 2012DLRM 2019Switch 2021Grok 3 2025

2012201620202024

Year

Arithmetic, not a forecast: if the 2018-to-2025 line (x2.02 a year) simply continued, the largest model would reach 150 trillion parameters around 2030. If the flatter 2021-to-2026 line continued, it would take well over a century. Neither is a prediction. Total parameter counts are not the only way models grow: longer training, more data and better methods raise ability without raising size, and the 2021-2026 stall in size coincides with the spread of sparse models and with labs ceasing to publish counts. Our study of AI and your brain looks at what people do with these systems, and whether AI can pass an IQ test at what they score.

10 Do a trillion parameters mean a trillion working connections?

No: of the 27 models with a trillion parameters or more, 20 are sparse or mixture-of-experts models, and the 10 largest of them whose notes state the active count use only 3% to 16% of their parameters for any one token ^{[1] [2]}. Epoch’s column holds the total, and its own guidance says that for a sparse model the compute heuristic should use the active experts. Kimi K3 stores 2.8 trillion parameters and uses 104 billion; MAI-Base-2 stores 2.8 trillion and uses 80 billion; GPT-4, if the rumour Epoch records is right, stores 1.8 trillion and uses 280 billion.

Figure 10

A trillion parameters stored, a few tens of billions used: the 10 biggest sparse models with a stated active count use 3% to 16%

Active parameters per token (orange) against total stored parameters (navy), in billions, for the 10 largest sparse models whose notes state both. The text and chart count the total, as Epoch's column does. GPT-4 is a rumoured figure (Epoch confidence "Likely"). As of 10 November 2026.

Active per tokenTotal stored

Kimi K3104B2800B

MAI-Base-280B2800B

Qwen 3.8 Max95B2400B

Llama 4 Behemoth (preview)288B2000B

GPT-4 (rumoured)280B1800B

DeepSeek-V4-Pro49B1600B

LongCat-2.048B1600B

Mistral Large 449B1050B

MiMo-V2-Pro42B1020B

Inkling41B975B

050010001500200025003000

Parameters in billions (B)

Read this way, the gap to a brain widens: these ten models use 41 to 288 billion parameters at a time, from half a mouse neocortex (81 billion synapses) to a little over three of them, and far under a cat’s 2.2 trillion. The brain has its own sparsity – most synapses are quiet at any moment – and no one knows how to compare the two. We therefore use the total, as the table does, and tell the reader what it hides. A fair version of the comparison would count active connections in both, which neither the models’ papers nor the brain’s counts supply.

11 Which AI companies publish how big their models are?

Epoch AI’s table has a parameter count for 41% of the notable models of 2026, against 92% in 2021, and for none of the 23 Anthropic models since 2023, 6 of 59 OpenAI models and 13 of 44 Google DeepMind models ^[1]. DeepSeek, which prints all 10 of its models’ sizes, and Alibaba, with 28 of 33, show that publishing a size is possible. Where a lab has not published one, the largest model in the world may be bigger than the 3 trillion on this page, and we cannot say by how much.

One row of the table carries an unconfirmed industry estimate of about 8 trillion parameters in its note while its parameter column is empty; we do not use it, because filling an empty cell with a rumour would turn a missing number into a false one. Even 8 trillion would sit 19 times below a human neocortex. GPT-4 is the best-known case: OpenAI has never published its size, and Epoch lists 1.8 trillion as a rumoured figure with a confidence of “Likely”, describing a mixture-of-experts model that uses about 280 billion parameters at once ^[1]. ChatGPT is a product built on such models; its size is not published either, and the row Epoch lists for GPT-3.5 Turbo gives 20 billion, taken from a Microsoft paper.

Figure 11

Fewer labs say how big their models are: a size was printed for 92% of notable models in 2021, 41% in 2026

Share of the notable models of each year that have a parameter count in Epoch AI's table, 2012 to 2026 (2026 runs to 28 October). Since 2023, OpenAI has 6 of 59 models with a count, Anthropic 0 of 23, Google DeepMind 13 of 44; DeepSeek 10 of 10 and Alibaba 28 of 33 print theirs. As of 10 November 2026.

69%'12

63%'13

64%'14

69%'15

61%'16

45%'17

81%'18

82%'19

86%'20

92%'21

79%'22

76%'23

57%'24

53%'25

41%'26

12 Who builds the biggest AI models?

Industry builds 112 of the 128 models at or above a mouse neocortex, and 60 of the 128 models at that size come from developers wholly in China, against 44 wholly in the United States ^[1]. This is a count of rows in a table by the country of the developing organisation, not a ranking of countries or of quality, and it depends on which models Epoch AI marks notable.

Figure 12

US and Chinese developers: 354 vs 117 single-country models, but China has more above a mouse

Notable models with a parameter count whose developing organisations are all in the United States (354 models) or all in China (117), by size tier. Above the mouse tier: United States 44, China 60. Models from several countries or other single countries are not shown. A count of models, not a measure of quality. As of 10 November 2026.

United States354 modelsTierChina117 models

3cat +3

41mouse-cat57

142bee-mouse40

72fly-bee11

96< fly6

50100150050100150

Figure 13

Who builds the big ones: industry makes 112 of the 128 models above a mouse

Width = number of notable models with a parameter count (728). Left: the developer's sector in Epoch's table (Industry, Academia, a mix of sectors or other, not stated). Right: the size tier. Industry makes 112 of the 128 models in the two top tiers; academia makes 2. As of 10 November 2026.

Developer sectorSectorSize tierTier

Industry (395)Industry 395Academia (156)Acad. 156Mixed or other (160)Mixed 160Not stated (17)n/s 17

Cat or morecat + 6Mouse to catmouse-cat 122Honeybee to mousebee-mouse 251Fruit fly to honeybeefly-bee 131Below a fruit fly< fly 218

Figure 14

Every model above a mouse, one square each: 60 from Chinese developers, 44 from US developers

One square = one of the 128 notable models with at least 81 billion parameters, coloured by the country of the developing organisation in Epoch's table: China 60, United States 44, another single country 18, several countries 6. Of the 60 Chinese models, 45 have open weights and 40 are described as sparse; of the 44 US models, 13 and 18. As of 10 November 2026.

China 60United States 44Other single country 18Several countries 6

128 modelsone square = one model

13 Does a bigger AI model cost more to train?

Yes: in the 478 notable models with both numbers, each tenfold rise in training compute goes with a 3.2-fold rise in parameters (a least-squares slope of 0.50 on log scales, correlation 0.92) ^[1]. Compute rises faster than size because models are also trained on more data. This is a link between two counts in one table, not a cause, and the compute figures are Epoch AI estimates of varying confidence.

Figure 15

Bigger models cost more to train: each tenfold rise in compute goes with a 3.2-fold rise in size (r = 0.92)

478 notable models with both a parameter count and a training-compute estimate: compute in floating-point operations across, parameters up, both as a power of ten, with a least-squares line (slope 0.50). A link, not a cause: compute also buys more training data. As of 10 November 2026.

Before 20182018 to 20222023 to 2026

Parameters as a power of ten

36912

510152025

Training compute, FLOP, as a power of ten

14 What does this comparison not show?

It does not show intelligence, ability or thinking: it compares how many adjustable links a model and a brain have, and size is not capacity. A honeybee, with about one million neurons, navigates and learns (see what bees, ants and flies can do), and models of any size can still fail puzzles that people solve on ARC-AGI-3. The link between brain size and human test scores is also weak, about 0.24 in the research on brain size and IQ. Our page on the smartest animals sets neuron counts for 222 species beside behavioural tests for some of them.

It does not forecast anything: the arithmetic line is a reading of a past trend. And it does not rest on exact numbers – three of the eleven animal rows are calculations, the Epoch AI column mixes confident and speculative values, and the animal ranges are 7 times wide for the fruit fly and 21 times for our honeybee bracket. The conclusion that survives every choice is the order of magnitude: the largest published model is about 1.7 orders of magnitude (50 to 55 times) below a human neocortex and about level with a cat.

15 Every animal and every AI model in our dataset

Table 1

Every animal in our dataset (11 brains and 2 samples)

Synapses = low and high count; "Counted" = a paper prints the number, "IQ Metrics calculation" = neurons times synapses per neuron (see the CSV). Mammals are neocortex only. First model = the first notable AI model at or above the low and the high count. As of 10 November 2026.

Animal	What is counted	Neurons	Synapses, low	Synapses, high	How obtained	First model at or above low	First model at or above high
Sea squirt larva (Ciona) [5]	whole central nervous system, tadpole larva	177	4,846	6,618	Counted	Cognitron (1975)	Cognitron (1975)
Roundworm (C. elegans) [3] [4]	whole nervous system, hermaphrodite	302	5,000	8,000	Counted	Cognitron (1975)	Cognitron (1975)
Marine worm larva (Platynereis, 3 days) [6]	whole body, 3-day larva	–	26,881	32,381	Counted	Neocognitron (1980)	Neocognitron (1980)
Fruit fly larva (first instar) [7]	whole brain, first-instar larva	3,013	544,000	544,000	Counted	Neocognitron (1980)	Neocognitron (1980)
Zebrafish larva (6 days) [8]	whole brain, retina and anterior spinal cord, algorithmic reconstruction	–	25,250,000	25,250,000	Counted	Support Vector Machines (1995)	Support Vector Machines (1995)
Fruit fly (adult) [9] [10]	whole brain (FlyWire, female); male central nervous system (Berg et al.)	139,255	46,000,000	312,000,000	Counted	Support Vector Machines (1995)	KN-LM (2007)
Honeybee [21] [22] [9] [13]	whole brain, estimate (no count published)	1,000,000	391,370,000	8,200,000,000	IQ Metrics calculation	KN-LM (2007)	KN-LM (2007)
Mouse (neocortex) [13]	neocortex, both hemispheres	10,000,000	81,000,000,000	81,000,000,000	Counted	SB-LM (2007)	SB-LM (2007)
Cat (neocortex) [19] [16]	cerebral cortex, calculated from neurons and synapses per neuron	249,830,000	2,239,000,000,000	2,439,000,000,000	IQ Metrics calculation	Grok 3 (2025)	Grok 3 (2025)
Macaque (neocortex) [20] [17]	cerebral cortex, calculated from neurons and synapses per neuron	800,960,000	3,123,744,000,000	6,474,000,000,000	IQ Metrics calculation	none yet	none yet
Human (neocortex) [15] [14] [18]	neocortex, both hemispheres	16,300,000,000	150,000,000,000,000	164,000,000,000,000	Counted	none yet	none yet
Mouse, 1 cubic millimetre of visual cortex (MICrONS) [11]	1 cubic millimetre of cortex, not a whole brain	–	524,000,000	524,000,000	Counted (automatic)	not compared	not compared
Human, 1 cubic millimetre of temporal cortex (H01) [12]	1 cubic millimetre of cortex, not a whole brain	–	133,700,000	133,700,000	Counted (automatic)	not compared	not compared

Table 2

The 25 largest notable models with a parameter count

Sorted by total parameters; ties share a size. Active = active parameters per token where the note states it. Confidence is Epoch AI's label for its estimate (Confident, Likely, Speculative, Unknown). Models with no count (for example every Anthropic model) are not ranked. As of 10 November 2026.

Rank	Model	Published	Developer	Parameters (total)	Active per token	Epoch confidence	Note
1	Grok 3	2025-02-17	xAI	3 trillion	–	Likely	–
2	Grok 4	2025-07-09	xAI	3 trillion	–	Speculative	rumoured
3	Kimi K3	2026-07-16	Moonshot	2.8 trillion	104 billion	Speculative	sparse (MoE), rumoured
4	MAI-Base-2	2026-10-15	Microsoft	2.8 trillion	80 billion	Confident	sparse (MoE)
5	Qwen 3.8 Max	2026-07-19	Alibaba	2.4 trillion	95 billion	Confident	sparse (MoE)
6	Qwen3.8-2.4T-A95B	2026-08-08	Alibaba	2.4 trillion	95 billion	Confident	sparse (MoE)
7	Grok 4.7	2026-09-21	xAI	2.1 trillion	–	Likely	rumoured
8	Llama 4 Behemoth (preview)	2025-04-05	Meta AI	2 trillion	288 billion	Likely	sparse (MoE)
9	GPT-4 (Mar 2023)	2023-03-15	OpenAI	1.8 trillion	280 billion	Likely	sparse (MoE), rumoured
10	GPT-4 (Jun 2023)	2023-06-13	OpenAI	1.8 trillion	280 billion	Likely	sparse (MoE), rumoured
11	DeepSeek-V4-Pro	2026-04-24	DeepSeek	1.6 trillion	49 billion	Likely	sparse (MoE)
12	LongCat-2.0	2026-06-29	Meituan Inc	1.6 trillion	48 billion	Likely	sparse (MoE)
13	Switch	2021-01-11	Google	1.6 trillion	–	Confident	–
14	GLaM	2021-12-13	Google	1.2 trillion	–	Confident	–
15	PanGu-Σ	2023-03-20	Huawei Noah’s Ark Lab	1.1 trillion	–	Confident	sparse (MoE)
16	Mistral Large 4	2026-10-06	Mistral AI	1.05 trillion	49 billion	Confident	sparse (MoE)
17	Kimi K2.5	2026-02-02	Moonshot	1 trillion	–	Likely	–
18	Composer 2	2026-03-19	Cursor	1 trillion	32 billion	Confident	sparse (MoE)
19	Kimi K2.6	2026-04-20	Moonshot	1 trillion	–	Confident	–
20	Composer 2.5	2026-05-18	Cursor	1 trillion	32 billion	Likely	sparse (MoE)
21	MiMo-V2-Pro	2026-03-18	Xiaomi Corp	1 trillion	42 billion	Confident	sparse (MoE)
22	MiMo-V2.5-Pro	2026-04-23	Xiaomi Corp	1 trillion	42 billion	Likely	sparse (MoE)
23	M6-T	2021-03-05	Alibaba	1 trillion	–	Likely	sparse (MoE)
24	Kimi K2	2025-07-11	Moonshot	1 trillion	–	Confident	sparse (MoE)
25	Qwen3-Max	2025-09-05	Alibaba	1 trillion	–	Speculative	sparse (MoE), rumoured

Table 3

Every notable AI model with a parameter count (728 rows)

Largest first; click a header to sort. Parameters as Epoch AI records them (total for sparse models). Tier = the animal the model has reached: < fly, fly-bee, bee-mouse, mouse-cat, cat +. The other 364 notable models have no count and are in the CSV with an empty size. As of 10 November 2026.

Model	Published	Developer	Country	Parameters	Tier	Epoch confidence	Note
Grok 3	2025-02-17	xAI	United States	3,000,000,000,000	cat +	Likely	–
Grok 4	2025-07-09	xAI	United States	3,000,000,000,000	cat +	Speculative	unconfirmed
Kimi K3	2026-07-16	Moonshot	China	2,800,000,000,000	cat +	Speculative	sparse (MoE), unconfirmed
MAI-Base-2	2026-10-15	Microsoft	United States	2,800,000,000,000	cat +	Confident	sparse (MoE)
Qwen 3.8 Max	2026-07-19	Alibaba	China	2,400,000,000,000	cat +	Confident	sparse (MoE)
Qwen3.8-2.4T-A95B	2026-08-08	Alibaba	China	2,400,000,000,000	cat +	Confident	sparse (MoE)
Grok 4.7	2026-09-21	xAI	United States	2,100,000,000,000	mouse-cat	Likely	unconfirmed
Llama 4 Behemoth (preview)	2025-04-05	Meta AI	United States	2,000,000,000,000	mouse-cat	Likely	sparse (MoE)
GPT-4 (Mar 2023)	2023-03-15	OpenAI	United States	1,800,000,000,000	mouse-cat	Likely	sparse (MoE), unconfirmed
GPT-4 (Jun 2023)	2023-06-13	OpenAI	United States	1,800,000,000,000	mouse-cat	Likely	sparse (MoE), unconfirmed
DeepSeek-V4-Pro	2026-04-24	DeepSeek	China	1,600,000,000,000	mouse-cat	Likely	sparse (MoE)
LongCat-2.0	2026-06-29	Meituan Inc	China	1,600,000,000,000	mouse-cat	Likely	sparse (MoE)
Switch	2021-01-11	Google	United States	1,571,000,000,000	mouse-cat	Confident	–
GLaM	2021-12-13	Google	United States	1,200,000,000,000	mouse-cat	Confident	–
PanGu-Σ	2023-03-20	Huawei Noah’s Ark Lab	China	1,085,000,000,000	mouse-cat	Confident	sparse (MoE)
Mistral Large 4	2026-10-06	Mistral AI	France	1,050,000,000,000	mouse-cat	Confident	sparse (MoE)
Kimi K2.5	2026-02-02	Moonshot	China	1,040,000,000,000	mouse-cat	Likely	–
Composer 2	2026-03-19	Cursor	United States	1,040,000,000,000	mouse-cat	Confident	sparse (MoE)
Kimi K2.6	2026-04-20	Moonshot	China	1,040,000,000,000	mouse-cat	Confident	–
Composer 2.5	2026-05-18	Cursor	United States	1,040,000,000,000	mouse-cat	Likely	sparse (MoE)
MiMo-V2-Pro	2026-03-18	Xiaomi Corp	China	1,020,000,000,000	mouse-cat	Confident	sparse (MoE)
MiMo-V2.5-Pro	2026-04-23	Xiaomi Corp	China	1,020,000,000,000	mouse-cat	Likely	sparse (MoE)
M6-T	2021-03-05	Alibaba	China	1,002,700,000,000	mouse-cat	Likely	sparse (MoE)
Kimi K2	2025-07-11	Moonshot	China	1,000,000,000,000	mouse-cat	Confident	sparse (MoE)
Qwen3-Max	2025-09-05	Alibaba	China	1,000,000,000,000	mouse-cat	Speculative	sparse (MoE), unconfirmed
Ling-1T	2025-10-10	Ant Group	China	1,000,000,000,000	mouse-cat	Confident	sparse (MoE)
Kimi K2 Thinking	2025-11-06	Moonshot	China	1,000,000,000,000	mouse-cat	Likely	sparse (MoE)
Inkling	2026-07-15	Thinking Machines	United States	975,000,000,000	mouse-cat	Likely	sparse (MoE)
Tencent Hy4 preview	2026-08-28	Tencent	China	770,000,000,000	mouse-cat	Likely	–
GLM-5.1	2026-04-07	Z.ai (Zhipu AI)	China	754,000,000,000	mouse-cat	Confident	–
K-EXAONE 2.0	2026-07-31	LG AI Research	South Korea	750,000,000,000	mouse-cat	Confident	sparse (MoE)
GLM-5	2026-02-12	Z.ai (Zhipu AI)	China	744,000,000,000	mouse-cat	Likely	–
GLM-5.3	2026-08-14	Z.ai (Zhipu AI)	China	744,000,000,000	mouse-cat	Confident	sparse (MoE)
A.X K2	2026-07-29	SK Telecom	South Korea	688,000,000,000	mouse-cat	Confident	sparse (MoE)
DeepSeekMath-V2	2025-11-27	DeepSeek	China	685,000,000,000	mouse-cat	Confident	–
DeepSeek-V3	2024-12-24	DeepSeek	China	671,000,000,000	mouse-cat	Confident	sparse (MoE)
DeepSeek-R1	2025-01-20	DeepSeek	China	671,000,000,000	mouse-cat	Confident	sparse (MoE)
DeepSeek-V3 (Mar 2025)	2025-03-24	DeepSeek	China	671,000,000,000	mouse-cat	Confident	sparse (MoE)
DeepSeek-R1 (May 2025)	2025-05-28	DeepSeek	China	671,000,000,000	mouse-cat	Confident	sparse (MoE)
SenseChat 5.5	2024-07-06	SenseTime	Hong Kong	600,000,000,000	mouse-cat	Confident	–
Step 5 Preview	2026-09-18	StepFun	China	600,000,000,000	mouse-cat	Likely	–
PaLM-E	2023-03-06	Google,TU Berlin	Germany; United States	562,000,000,000	mouse-cat	Likely	–
Hunyuan-TurboS	2025-03-11	Tencent	China	560,000,000,000	mouse-cat	Confident	sparse (MoE)
LongCat-Flash	2025-09-01	Meituan Inc	China	560,000,000,000	mouse-cat	Likely	sparse (MoE)
DeepSeek V4.1 Flash	2026-09-09	DeepSeek	China	552,000,000,000	mouse-cat	Likely	sparse (MoE)
Nemotron 3 Ultra	2026-06-04	Nvidia	United States	550,000,000,000	mouse-cat	Likely	sparse (MoE)
PaLM (540B)	2022-04-04	Google Research	United States	540,350,000,000	mouse-cat	Confident	sparse (MoE)
Minerva (540B)	2022-06-29	Google	United States	540,350,000,000	mouse-cat	Confident	sparse (MoE)
U-PaLM (540B)	2022-10-20	Google	United States	540,000,000,000	mouse-cat	Confident	–
LMSI-Palm	2022-10-20	Google,University of Illinois Urbana-Champaign (UI	United States	540,000,000,000	mouse-cat	Confident	–
Flan-PaLM 540B	2022-10-20	Google	United States	540,000,000,000	mouse-cat	Confident	–
Megatron-Turing NLG 530B	2021-10-11	Microsoft,Nvidia	United States	530,000,000,000	mouse-cat	Confident	–
MegaScale (Production)	2024-02-23	ByteDance,Peking University	China	530,000,000,000	mouse-cat	Speculative	unconfirmed
A.X K1	2025-12-30	SK Telecom	South Korea	519,000,000,000	mouse-cat	Confident	sparse (MoE)
Doubao-pro	2024-10-28	ByteDance	China	500,000,000,000	mouse-cat	Likely	–
Grok 4.20	2026-02-17	xAI	United States	500,000,000,000	mouse-cat	Likely	sparse (MoE)
Qwen3-Coder-480B-A35B	2025-07-22	Alibaba	China	480,000,000,000	mouse-cat	Confident	sparse (MoE)
MiniMax-M3	2026-06-01	MiniMax	China	428,000,000,000	mouse-cat	Confident	sparse (MoE)
ERNIE-4.5-VL-424B-A47B (文心大模型4.5)	2025-03-16	Baidu	China	424,000,000,000	mouse-cat	Confident	sparse (MoE)
Llama 3.1-405B	2024-07-23	Meta AI	United States	405,000,000,000	mouse-cat	Confident	–
Llama 4 Maverick	2025-04-05	Meta AI	United States	400,000,000,000	mouse-cat	Likely	sparse (MoE)
Jamba 1.5-Large	2024-08-22	AI21 Labs	Israel	398,000,000,000	mouse-cat	Confident	sparse (MoE)
Qwen3.5 397B-A17B	2026-02-15	Alibaba	China	397,000,000,000	mouse-cat	Likely	–
Hunyuan-Large	2024-11-06	Tencent	China	389,000,000,000	mouse-cat	Confident	–
GLM-4.7	2025-12-22	Z.ai (Zhipu AI)	China	358,000,000,000	mouse-cat	Likely	sparse (MoE)
GLM-4.6	2025-09-30	Z.ai (Zhipu AI),Tsinghua University	China	357,000,000,000	mouse-cat	Likely	sparse (MoE)
GLM-4.5	2025-07-28	Z.ai (Zhipu AI),Tsinghua University	China	355,000,000,000	mouse-cat	Confident	sparse (MoE)
PaLM 2	2023-05-10	Google	United States	340,000,000,000	mouse-cat	Likely	–
Med-PaLM 2	2023-05-16	Google Research,DeepMind	United Kingdom; United States	340,000,000,000	mouse-cat	Likely	–
Nemotron-4 340B	2024-06-14	Nvidia	United States	340,000,000,000	mouse-cat	Confident	–
GLM-5.3-Flash	2026-08-20	Z.ai (Zhipu AI)	China	320,000,000,000	mouse-cat	Likely	sparse (MoE)
Grok-1	2023-11-04	xAI	United States	314,000,000,000	mouse-cat	Likely	sparse (MoE)
Motif-3	2026-08-07	Motif Technologies	South Korea	314,000,000,000	mouse-cat	Confident	sparse (MoE)
SB-LM	2007-06-22	Google	United States	300,000,000,000	mouse-cat	Likely	not a neural network
Tencent Hy3 preview	2026-04-23	Tencent	China	295,000,000,000	mouse-cat	Confident	sparse (MoE)
Tencent Hy3	2026-07-06	Tencent	China	295,000,000,000	mouse-cat	Confident	sparse (MoE)
DeepSeek-V4-Flash	2026-04-24	DeepSeek	China	284,000,000,000	mouse-cat	Likely	sparse (MoE)
Gopher (280B)	2021-12-08	DeepMind	United Kingdom	280,000,000,000	mouse-cat	Confident	–
TML-Interaction-Small	2026-05-11	Thinking Machines	United States	276,000,000,000	mouse-cat	Confident	sparse (MoE)
ST-MoE	2022-02-17	Google,Google Brain,Google Research	United States	269,000,000,000	mouse-cat	Likely	sparse (MoE)
ERNIE 3.0 Titan	2021-12-23	Baidu,Peng Cheng Laboratory	China	260,000,000,000	mouse-cat	Confident	–
Solar Open2 250B	2026-06-28	Upstage	South Korea	250,300,000,000	mouse-cat	Likely	sparse (MoE)
Aramco Metabrain AI	2024-03-04	Saudi Aramco	Saudi Arabia	250,000,000,000	mouse-cat	Likely	–
Yuan 1.0	2021-10-12	Inspur	China	245,730,000,000	mouse-cat	Confident	–
DeepSeek-Coder-V2 236B	2024-06-17	DeepSeek	China	236,000,000,000	mouse-cat	Confident	sparse (MoE)
DeepSeek-V2.5	2024-09-06	DeepSeek	China	236,000,000,000	mouse-cat	Confident	sparse (MoE)
K-EXAONE	2025-12-31	LG AI Research	South Korea	236,000,000,000	mouse-cat	Confident	–
Qwen3-235B-A22B	2025-04-29	Alibaba	China	235,000,000,000	mouse-cat	Likely	sparse (MoE)
Qwen3-235B-A22B-Thinking (Jul 2025)	2025-07-25	Alibaba	China	235,000,000,000	mouse-cat	Likely	sparse (MoE)
Qwen3-235B-A22B (Jul 2025)	2025-07-25	Alibaba	China	235,000,000,000	mouse-cat	Likely	sparse (MoE)
Qwen3-VL-235B-A22B	2025-09-23	Alibaba	China	235,000,000,000	mouse-cat	Likely	sparse (MoE)
P1-235B-A22B	2025-11-17	Shanghai AI Lab	China	235,000,000,000	mouse-cat	Likely	sparse (MoE)
Seed-1.6-Thinking	2025-06-11	ByteDance	China	230,000,000,000	mouse-cat	Confident	sparse (MoE)
MiniMax-M2	2025-10-27	MiniMax	China	229,000,000,000	mouse-cat	Confident	–
MiniMax-M2.1	2025-12-23	MiniMax	China	229,000,000,000	mouse-cat	Confident	sparse (MoE)
HyperCLOVA 204B	2021-09-10	NAVER	South Korea	204,000,000,000	mouse-cat	Speculative	unconfirmed
Amazon Titan	2023-09-28	Amazon	United States	200,000,000,000	mouse-cat	Likely	–
Falcon-180B	2023-09-06	Technology Innovation Institute	United Arab Emirates	180,000,000,000	mouse-cat	Confident	–
Jurassic-1-Jumbo	2021-08-11	AI21 Labs	Israel	178,000,000,000	mouse-cat	Confident	–
BLOOM-176B	2022-07-11	Hugging Face,BigScience	France; United States	176,247,271,424	mouse-cat	Confident	–
BLUUMI	2023-11-03	University of Turku,Hugging Face	Finland; United States	176,000,000,000	mouse-cat	Likely	–
InstructGPT 175B	2022-01-27	OpenAI	United States	175,000,000,000	mouse-cat	Confident	–
OPT-175B	2022-05-02	Meta AI	United States	175,000,000,000	mouse-cat	Confident	–
BlenderBot 3	2022-08-10	McGill University,Meta AI,Mila – Quebec AI (origin	Canada; United States	175,000,000,000	mouse-cat	Likely	–
GPT-3 175B (davinci)	2020-05-28	OpenAI	United States	174,600,000,000	mouse-cat	Confident	–
Palmyra X 004	2024-10-09	Writer	United States	150,000,000,000	mouse-cat	–	–
FLAN 137B	2021-09-03	Google Research	United States	137,000,000,000	mouse-cat	Confident	–
LaMDA	2022-02-10	Google	United States	137,000,000,000	mouse-cat	Confident	–
MEB	2021-09-04	Microsoft	United States	135,000,000,000	mouse-cat	Confident	–
Pangu Ultra	2025-04-10	Huawei	China	135,000,000,000	mouse-cat	Confident	–
DBRX	2024-03-27	Databricks	United States	132,000,000,000	mouse-cat	Confident	sparse (MoE)
GLM-130B	2022-08-04	Tsinghua University	China	130,000,000,000	mouse-cat	Confident	–
Pixtral Large	2024-11-18	Mistral AI	France	124,000,000,000	mouse-cat	Confident	–
Mistral Large 2	2024-07-24	Mistral AI	France	123,000,000,000	mouse-cat	Likely	–
Qwen3.5-122B-A10B	2026-02-24	Alibaba	China	122,000,000,000	mouse-cat	Likely	–
Galactica	2022-11-16	Meta AI	United States	120,000,000,000	mouse-cat	Likely	–
Nemotron 3 Super	2026-03-11	Nvidia	United States	120,000,000,000	mouse-cat	Likely	sparse (MoE)
gpt-oss-120b	2025-08-05	OpenAI	United States	116,830,000,000	mouse-cat	Confident	–
Telechat2-115B	2024-09-20	China Telecom	China	115,000,000,000	mouse-cat	Confident	–
Kaiju Large	2025-11-07	Character.ai	United States	110,000,000,000	mouse-cat	Likely	–
Llama 4 Scout	2025-04-05	Meta AI	United States	109,000,000,000	mouse-cat	Likely	sparse (MoE)
InternLM	2023-07-06	Shanghai AI Lab,SenseTime	China; Hong Kong	104,000,000,000	mouse-cat	Confident	–
Solar Open 100B	2025-12-31	Upstage	South Korea	102,000,000,000	mouse-cat	Confident	sparse (MoE)
DLRM-2020	2019-05-31	Facebook AI	United States	100,000,000,000	mouse-cat	Confident	–
xTrimoPGLM -100B	2023-07-06	Tsinghua University,BioMap Research	China	100,000,000,000	mouse-cat	Confident	–
Yi-Large	2024-05-13	01.AI	China	100,000,000,000	mouse-cat	Speculative	unconfirmed
VAETKI	2025-12-30	NC AI	South Korea	100,000,000,000	mouse-cat	Confident	sparse (MoE)
ESM3 (98B)	2024-06-25	EvolutionaryScale,University of California (UC) Be	United States	98,500,000,000	mouse-cat	Confident	–
Flamingo	2022-04-29	DeepMind	United Kingdom	80,000,000,000	bee-mouse	Confident	–
Qwen3-Coder-Next	2026-02-02	Alibaba	China	80,000,000,000	bee-mouse	Likely	–
Qwen2-72B	2024-06-07	Alibaba	China	72,710,000,000	bee-mouse	Confident	–
Qwen2.5-72B	2024-09-19	Alibaba	China	72,700,000,000	bee-mouse	Confident	–
Qwen2.5 Instruct (72B)	2024-09-19	Alibaba	China	72,700,000,000	bee-mouse	Confident	–
Qwen-72B	2023-11-30	Alibaba	China	72,000,000,000	bee-mouse	Confident	–
Palmyra X 003	2024-01-01	Writer	United States	72,000,000,000	bee-mouse	–	–
Qwen1.5-72B	2024-02-04	Alibaba	China	72,000,000,000	bee-mouse	Confident	–
LLaVA-OV-72B	2024-08-06	ByteDance,Nanyang Technological University,Chinese	China; Hong Kong; Singapore	72,000,000,000	bee-mouse	Confident	–
NVLM-X 72B	2024-10-22	Nvidia	United States	72,000,000,000	bee-mouse	Likely	–
NVLM-H 72B	2024-10-22	Nvidia	United States	72,000,000,000	bee-mouse	Likely	–
NVLM-D 72B	2024-10-22	Nvidia	United States	72,000,000,000	bee-mouse	Confident	–
MindLink-72B	2025-08-01	Kunlun Inc.	China	72,000,000,000	bee-mouse	Confident	–
Chinchilla	2022-03-29	DeepMind	United Kingdom	70,000,000,000	bee-mouse	Confident	–
Llama 2-70B	2023-07-18	Meta AI	United States	70,000,000,000	bee-mouse	Confident	–
PPLX-70B-Online	2023-11-29	Perplexity	United States	70,000,000,000	bee-mouse	Likely	–
Llama 3-70B	2024-04-18	Meta AI	United States	70,000,000,000	bee-mouse	Confident	–
ALLaM adapted 70B	2024-05-21	Saudi Data and Artificial Intelligence Authority	Saudi Arabia	70,000,000,000	bee-mouse	Confident	–
Llama 3.3 70B	2024-12-06	Meta AI	United States	70,000,000,000	bee-mouse	Confident	–
Reka Core	2024-04-15	Reka AI	United States	67,000,000,000	bee-mouse	Speculative	unconfirmed
LLaMA-65B	2023-02-24	Meta AI	United States	65,200,000,000	bee-mouse	Confident	–
SNM-skip	2014-12-03	Google	United States	62,000,000,000	bee-mouse	Speculative	unconfirmed
PaLI-X	2023-05-29	Google Research	United States	55,000,000,000	bee-mouse	Likely	–
RT-2	2023-07-28	Google DeepMind	United States	55,000,000,000	bee-mouse	Confident	–
RT-2-X	2023-10-13	Google DeepMind	United States	55,000,000,000	bee-mouse	Confident	–
NLLB	2022-07-06	Meta AI	United States	54,500,000,000	bee-mouse	Confident	–
BloombergGPT	2023-03-30	Bloomberg,Johns Hopkins University	United States	50,558,868,480	bee-mouse	Confident	–
M4-50B	2019-10-11	Google	United States	50,000,000,000	bee-mouse	Confident	sparse (MoE)
Mixtral 8x7B	2023-12-11	Mistral AI	France	46,700,000,000	bee-mouse	Speculative	sparse (MoE), unconfirmed
AlphaCode	2022-02-02	DeepMind	United Kingdom	41,100,000,000	bee-mouse	Confident	–
Falcon-40B	2023-03-15	Technology Innovation Institute	United Arab Emirates	40,000,000,000	bee-mouse	Confident	–
Qwen3-Omni-30B-A3B	2025-09-22	Alibaba	China	35,300,000,000	bee-mouse	Confident	sparse (MoE)
Yi-34B	2023-11-02	01.AI	China	34,000,000,000	bee-mouse	Confident	–
Cambrian-1-34B	2024-06-24	New York University (NYU)	United States	34,000,000,000	bee-mouse	Confident	–
Oryx 34B	2024-09-19	Tsinghua University,Tencent,Nanyang Technological	China; Singapore	34,000,000,000	bee-mouse	Confident	–
Kaiju Medium	2025-11-07	Character.ai	United States	34,000,000,000	bee-mouse	Likely	–
EXAONE 4.5	2026-04-09	LG AI Research	South Korea	33,200,000,000	bee-mouse	Confident	–
Qwen2.5-32B	2024-09-17	Alibaba	China	32,500,000,000	bee-mouse	Confident	–
QwQ-32B	2025-03-06	Alibaba	China	32,500,000,000	bee-mouse	Speculative	unconfirmed
EXAONE 3.5 32B	2024-12-09	LG AI Research	South Korea	32,000,000,000	bee-mouse	Confident	–
EXAONE Deep 32B	2025-03-16	LG AI Research	South Korea	32,000,000,000	bee-mouse	Confident	–
EXAONE 4.0 (32B)	2025-07-15	LG AI Research	South Korea	32,000,000,000	bee-mouse	Confident	–
Olmo 3	2025-11-20	Allen Institute for AI	United States	32,000,000,000	bee-mouse	Confident	–
HyperCLOVA X SEED 32B Think	2025-12-29	NAVER	South Korea	32,000,000,000	bee-mouse	Confident	–
Nemotron 3-Nano-30B-A3B	2025-12-15	Nvidia	United States	31,600,000,000	bee-mouse	Likely	–
Tongyi DeepResearch	2025-10-28	Alibaba	China	30,500,000,000	bee-mouse	Confident	sparse (MoE)
MM1-30B	2024-03-14	Apple	United States	30,000,000,000	bee-mouse	Likely	–
Movie Gen Video	2024-10-04	Meta AI	United States	30,000,000,000	bee-mouse	Confident	–
AgentFounder-30B	2025-09-16	Alibaba	China	30,000,000,000	bee-mouse	Confident	–
PLUG	2021-04-19	Alibaba	China	27,000,000,000	bee-mouse	Confident	–
Qwen Image	2025-08-04	Alibaba	China	27,000,000,000	bee-mouse	Confident	–
GenSLM	2022-10-11	University of Chicago,Nvidia,Harvard University,Ce	Germany; United States	25,000,000,000	bee-mouse	Confident	–
ViT-22B	2023-02-10	Google	United States	21,743,000,000	bee-mouse	Confident	–
KN-LM	2007-06-22	Google	United States	21,000,000,000	bee-mouse	Likely	not a neural network
gpt-oss-20b	2025-08-05	OpenAI	United States	20,910,000,000	bee-mouse	Confident	–
RNN for 1B words	2013-12-11	Google	United States	20,000,000,000	bee-mouse	Speculative	unconfirmed
GPT-NeoX-20B	2022-02-09	EleutherAI	United States	20,000,000,000	bee-mouse	Confident	–
UL2	2022-05-10	Google Research,Google Brain	United States	20,000,000,000	bee-mouse	Confident	–
Parti	2022-06-22	Google Research	United States	20,000,000,000	bee-mouse	Confident	–
GPT-3.5 Turbo	2023-06-13	OpenAI	United States	20,000,000,000	bee-mouse	Likely	–
GPT-3.5 Turbo Instruct	2023-09-28	OpenAI	United States	20,000,000,000	bee-mouse	Likely	–
SPHINX (Llama 2 13B)	2023-11-13	Shanghai AI Lab,Chinese University of Hong Kong (C	China; Hong Kong	19,900,000,000	bee-mouse	Likely	–
AlexaTM 20B	2022-08-02	Amazon	United States	19,750,000,000	bee-mouse	Confident	–
CogAgent	2023-12-14	Tsinghua University,Z.ai (Zhipu AI)	China	18,000,000,000	bee-mouse	Likely	–
Turing-NLG	2020-02-13	Microsoft	United States	17,000,000,000	bee-mouse	Likely	–
CogVLM-17B	2023-11-06	Tsinghua University,Z.ai (Zhipu AI),Beihang Univer	China	17,000,000,000	bee-mouse	Confident	–
PaLI	2022-09-14	Google	United States	16,900,000,000	bee-mouse	Likely	–
CodeT5+	2023-05-20	Salesforce	United States	16,000,000,000	bee-mouse	Confident	–
StarCoder	2023-05-09	Hugging Face,ServiceNow,Northeastern University,Mi	Australia; Canada; Germany; India; Israe	15,500,000,000	bee-mouse	Confident	–
ESM2-15B	2022-07-21	Meta AI,New York University (NYU),Stanford Univers	United States	15,000,000,000	bee-mouse	Confident	–
FunSearch	2023-12-14	Google DeepMind	United States	15,000,000,000	bee-mouse	Speculative	unconfirmed
NVILA 15B	2024-12-05	Nvidia,Massachusetts Institute of Technology (MIT)	China; United States	15,000,000,000	bee-mouse	Confident	–
nekomata-14b	2023-12-21	rinna	Japan	14,200,000,000	bee-mouse	Confident	–
VILA1.5-13B	2024-05-03	Nvidia,Massachusetts Institute of Technology (MIT)	United States	13,493,916,736	bee-mouse	Confident	–
VILA-13B	2023-12-12	Nvidia,Massachusetts Institute of Technology (MIT)	United States	13,350,839,296	bee-mouse	Confident	–
mT5-XXL	2020-10-20	Google,Google Research	United States	13,000,000,000	bee-mouse	Confident	–
mT0-13B	2022-11-03	Hugging Face,BigScience	France; United States	13,000,000,000	bee-mouse	Confident	–
LLaVA	2023-04-17	University of Wisconsin Madison,Microsoft Research	United States	13,000,000,000	bee-mouse	Confident	–
InstructBLIP	2023-05-11	Salesforce Research,Hong Kong University of Scienc	Hong Kong; Singapore; United States	13,000,000,000	bee-mouse	Confident	–
Jais	2023-08-29	Cerebras Systems,Mohamed bin Zayed University of A	United Arab Emirates; United States	13,000,000,000	bee-mouse	Confident	–
FinGPT-13B	2023-10-07	University of California Los Angeles (UCLA),Columb	United States	13,000,000,000	bee-mouse	Likely	–
Ferret (13B)	2023-10-11	Columbia University,Apple	United States	13,000,000,000	bee-mouse	Confident	–
Skywork-13B	2023-10-30	Kunlun Inc.	China	13,000,000,000	bee-mouse	Confident	–
LLaVA 1.5	2023-11-05	University of Wisconsin Madison,Microsoft Research	United States	13,000,000,000	bee-mouse	Confident	–
Volcano 13B	2023-11-13	Korea University,Korea Advanced Institute of Scien	South Korea	13,000,000,000	bee-mouse	Likely	–
Aya	2024-02-12	Cohere Labs (formerly Cohere for AI),Brown Univers	Canada; United States	13,000,000,000	bee-mouse	Speculative	unconfirmed
Kaiju Small	2025-11-07	Character.ai	United States	13,000,000,000	bee-mouse	Likely	–
ByT5-XXL	2021-05-28	Google,Google Research	United States	12,900,000,000	bee-mouse	Likely	–
CoRe	2023-12-29	Tsinghua University	China	12,400,000,000	bee-mouse	Speculative	unconfirmed
DALL-E	2021-01-05	OpenAI	United States	12,000,000,000	bee-mouse	Likely	–
Codex	2021-07-07	OpenAI	United States	12,000,000,000	bee-mouse	Likely	–
T5-11B	2019-10-23	Google	United States	11,000,000,000	bee-mouse	Confident	–
UnifiedQA	2020-05-02	Allen Institute for AI,University of Washington	United States	11,000,000,000	bee-mouse	Confident	–
Unicorn	2021-03-24	Allen Institute for AI	United States	11,000,000,000	bee-mouse	Confident	–
PLATO-XL	2021-09-20	Baidu	China	11,000,000,000	bee-mouse	Confident	–
CoEdiT-xxl	2023-05-17	University of Minnesota,Grammarly	United States	11,000,000,000	bee-mouse	Likely	–
GQA-8-XXL	2023-12-23	Google Research	United States	11,000,000,000	bee-mouse	Confident	–
XLMR-XXL	2021-08-17	Facebook AI Research	France; United States	10,700,000,000	bee-mouse	Confident	–
Llama 3.2 11B	2024-09-24	Meta AI	United States	10,600,000,000	bee-mouse	Confident	–
ERNIE 3.0	2021-07-05	Baidu	China	10,000,000,000	bee-mouse	Confident	–
ERNIE-ViLG	2021-12-31	Baidu	China	10,000,000,000	bee-mouse	–	–
Qwen-VL	2023-08-24	Alibaba	China	9,600,000,000	bee-mouse	Likely	–
Generative BST	2021-03-05	Facebook AI Research	France; United States	9,431,810,048	bee-mouse	Confident	–
Sparse all-MLP	2022-04-14	Meta AI	United States	9,410,000,000	bee-mouse	Confident	–
CogVideo	2022-05-29	Tsinghua University,Beijing Academy of Artificial	China	9,400,000,000	bee-mouse	Speculative	unconfirmed
eDiff-I	2022-11-02	Nvidia	United States	9,100,000,000	bee-mouse	Likely	–
Eagle 2	2025-01-20	Nvidia,Nanjing University,Tsinghua University,Hong	China; Hong Kong; United States	8,930,000,000	bee-mouse	Confident	–
MoE-Multi	2017-01-23	Jagiellonian University,Google Brain	Poland; United States	8,700,000,000	bee-mouse	Confident	–
Qwen-Audio-Chat	2023-11-14	Alibaba	China	8,460,000,000	bee-mouse	Likely	–
Megatron-LM (8.3B)	2019-09-17	Nvidia	United States	8,300,000,000	bee-mouse	Likely	–
BigSSL	2021-01-10	Google,Apple	United States	8,000,000,000	bee-mouse	–	–
Nemotron-3-8B	2023-11-15	Nvidia	United States	8,000,000,000	bee-mouse	Confident	–
Stable Diffusion 3	2024-02-22	Stability AI	United Kingdom	8,000,000,000	bee-mouse	Confident	–
Qwen3 Embedding	2025-06-05	Alibaba	China	8,000,000,000	bee-mouse	Confident	–
Imagen	2022-05-23	Google Brain	United States	7,762,000,000	bee-mouse	Likely	–
XGLM-7.5B	2021-12-20	Meta AI,Facebook AI Research	France; United States	7,500,000,000	bee-mouse	Confident	–
RETRO-7B	2022-02-07	DeepMind	United Kingdom	7,500,000,000	bee-mouse	Confident	–
OpenVLA	2024-06-13	Stanford University,University of California (UC)	United States	7,188,100,000	bee-mouse	Confident	–
mPLUG-Owl2	2023-11-07	Alibaba	China	7,120,000,000	bee-mouse	Speculative	unconfirmed
Goat-7B	2023-05-23	National University of Singapore	Singapore	7,000,000,000	bee-mouse	Speculative	unconfirmed
Llama 2-7B	2023-07-18	Meta AI	United States	7,000,000,000	bee-mouse	Confident	–
Llama Guard	2023-12-07	Meta AI	United States	7,000,000,000	bee-mouse	Confident	–
Qwen-VL-Max	2024-01-25	Alibaba	China	7,000,000,000	bee-mouse	Confident	–
Mathstral	2024-07-16	Mistral AI	France	7,000,000,000	bee-mouse	Confident	–
Apollo 7B	2024-12-13	Meta AI,Stanford University	United States	7,000,000,000	bee-mouse	Confident	–
Eurus-2-7B-PRIME	2025-01-02	Tsinghua University,University of Illinois Urbana-	China; United States	7,000,000,000	bee-mouse	Speculative	unconfirmed
INTELLECT-MATH	2025-01-17	Prime Intellect	United States	7,000,000,000	bee-mouse	Confident	–
Incoder-6.7B	2023-04-09	Facebook AI Research,University of Washington,Univ	France; United States	6,700,000,000	bee-mouse	Confident	–
ProGen2-xlarge	2022-06-27	Salesforce Research,Columbia University,Johns Hopk	United States	6,400,000,000	bee-mouse	Confident	–
GPipe (Transformer)	2018-11-16	Google	United States	6,000,000,000	bee-mouse	Likely	–
InstructGPT 6B	2022-01-27	OpenAI	United States	6,000,000,000	bee-mouse	Confident	–
Vega v2	2022-12-04	Wuhan University,JD Explore Academy,Shanghai AI La	Australia; China; Singapore; United Stat	6,000,000,000	bee-mouse	Confident	–
ChatGLM3-6B	2023-10-27	Z.ai (Zhipu AI)	China	6,000,000,000	bee-mouse	Speculative	unconfirmed
MolmoAct 2	2026-04-05	Allen Institute for AI,University of Washington,Na	Singapore; United States	5,485,309,424	bee-mouse	Likely	–
T-NLRv5 XXL	2021-12-03	Microsoft	United States	5,400,000,000	bee-mouse	Confident	–
π0.6 (pi-0.6)	2025-11-17	Physical Intelligence	United States	5,260,000,000	bee-mouse	Confident	–
PaLI-3	2023-10-17	Google DeepMind,Google Research,Google Cloud	United States	5,000,000,000	bee-mouse	Confident	–
Mesh-TensorFlow Transformer 4.9B (language)	2018-11-05	Google Brain	United States	4,900,000,000	bee-mouse	Confident	–
W.A.L.T	2023-12-11	Stanford University,Google Research,Georgia Instit	United States	4,719,000,000	bee-mouse	Confident	–
CogView	2021-05-26	Tsinghua University,Alibaba DAMO Academy	China	4,000,000,000	bee-mouse	Likely	–
Make-A-Scene	2022-03-24	Meta AI	United States	4,000,000,000	bee-mouse	Likely	–
ONE-PEACE	2023-05-18	Alibaba,Huazhong University of Science and Technol	China	4,000,000,000	bee-mouse	Speculative	unconfirmed
Megatron-BERT	2019-09-17	Nvidia	United States	3,900,000,000	bee-mouse	Confident	–
DALL·E 2	2022-04-06	OpenAI	United States	3,500,000,000	bee-mouse	Speculative	unconfirmed
Stable Diffusion XL (SDXL)	2023-07-04	Stability AI	United Kingdom	3,400,000,000	bee-mouse	Speculative	unconfirmed
MusicGen	2023-06-08	Meta AI	United States	3,359,000,000	bee-mouse	Likely	–
Gemini Nano-2	2023-12-19	Google DeepMind	United States	3,250,000,000	bee-mouse	Confident	–
Zidong Taichu	2021-08-11	Chinese Academy of Sciences,Wuhan AI Computing Cen	China	3,200,000,000	bee-mouse	Confident	–
DeepNet	2022-03-01	Microsoft Research	United States	3,200,000,000	bee-mouse	Confident	–
MS-ensemble-speech-recognition	2016-09-12	Microsoft	United States	3,172,117,056	bee-mouse	Likely	–
BASIC-L	2021-11-19	Google	United States	3,070,000,000	bee-mouse	Likely	–
BASIC-L + Lion	2023-02-13	Google,University of California Los Angeles (UCLA)	United States	3,070,000,000	bee-mouse	Confident	–
LEP-AD	2023-03-15	King Abdullah University of Science and Technology	Saudi Arabia; Sweden	3,007,381,000	bee-mouse	Confident	–
ViT-G/14 (LiT)	2021-11-15	Google Research	United States	3,005,000,000	bee-mouse	Confident	–
ProtT5-XL-U50	2021-05-04	Technical University of Munich,Med AI Technology,N	China; Germany; South Korea; United Stat	3,000,000,000	bee-mouse	Confident	–
Swin Transformer V2 (SwinV2-G)	2021-11-18	Microsoft Research Asia	China	3,000,000,000	bee-mouse	Confident	–
LongT5	2021-12-15	Google Research	United States	3,000,000,000	bee-mouse	Confident	–
ReALM	2024-03-29	Apple	United States	3,000,000,000	bee-mouse	Confident	–
Mesh-TensorFlow Transformer 2.9B (translation)	2018-11-05	Google Brain	United States	2,900,000,000	bee-mouse	Likely	–
T5-3B	2019-10-23	Google	United States	2,800,000,000	bee-mouse	Confident	–
DPO on Pythia-2.8B	2023-05-29	Stanford University,CZ Biohub Network	United States	2,800,000,000	bee-mouse	Confident	–
AFM-on-device	2024-07-29	Apple	United States	2,730,000,000	bee-mouse	Confident	–
PolyCoder	2022-02-26	Carnegie Mellon University (CMU)	United States	2,700,000,000	bee-mouse	Likely	–
Hybrid H3-2.7B	2022-12-28	Stanford University,University at Buffalo	United States	2,700,000,000	bee-mouse	Likely	–
GPT3-2.7B (FlashAttention-2)	2023-07-18	Stanford University,Princeton University	United States	2,700,000,000	bee-mouse	Confident	–
Meena	2020-01-28	Google Brain	United States	2,600,000,000	bee-mouse	Confident	–
CPM-Large	2020-12-01	Tsinghua University,Beijing Academy of Artificial	China	2,600,000,000	bee-mouse	Confident	–
Nucleotide Transformer	2023-01-15	Nvidia,Technical University of Munich,InstaDeep	Germany; United Kingdom; United States	2,500,000,000	bee-mouse	Likely	–
Fugatto 1	2024-11-25	Nvidia	United States	2,500,000,000	bee-mouse	Confident	–
CoAtNet	2021-06-09	Google,Google Research,Google Brain	United States	2,440,000,000	bee-mouse	Confident	–
GShard (dense)	2020-06-30	Google	United States	2,300,000,000	bee-mouse	Confident	–
SeamlessM4T	2023-12-08	Facebook,INRIA,University of California (UC) Berke	France; United States	2,300,000,000	bee-mouse	Confident	–
Turing ULRv5	2021-09-28	Microsoft	United States	2,200,000,000	bee-mouse	Confident	–
CoCa	2022-06-14	Google Research	United States	2,100,000,000	bee-mouse	Confident	–
MuseNet	2019-04-25	OpenAI	United States	2,038,431,744	bee-mouse	Likely	–
Infinity	2024-12-05	ByteDance	China	2,000,000,000	bee-mouse	Confident	–
Seq2Seq LSTM	2014-09-10	Google	United States	1,920,000,000	bee-mouse	Confident	–
BEIT-3	2022-08-22	Microsoft	United States	1,900,000,000	bee-mouse	Likely	–
Ankh_large	2023-01-16	Technical University of Munich,Columbia University	Germany; United States	1,900,000,000	bee-mouse	Confident	–
ViT-G/14	2021-06-08	Google Brain,Google Research	United States	1,843,000,000	bee-mouse	Confident	–
ViT-G (model soup)	2022-03-10	University of Washington,Columbia University,Googl	Israel; United States	1,843,000,000	bee-mouse	Confident	–
Phenaki	2022-10-05	University College London (UCL),University of Mich	United Kingdom; United States	1,800,000,000	bee-mouse	–	–
Gemini Nano-1	2023-12-19	Google DeepMind	United States	1,800,000,000	bee-mouse	Confident	–
DistBelief Vision	2012-12-03	Google	United States	1,700,000,000	bee-mouse	Likely	–
Whisper	2022-09-21	OpenAI	United States	1,550,000,000	bee-mouse	Likely	–
GPT-2 (1.5B)	2019-02-14	OpenAI	United States	1,500,000,000	bee-mouse	Speculative	unconfirmed
DeBERTa	2021-06-10	Microsoft	United States	1,500,000,000	bee-mouse	Confident	–
AR-LDM	2022-11-20	Alibaba,University of Waterloo,Vector Institute	Canada; China	1,500,000,000	bee-mouse	Confident	–
AudioLM	2023-07-26	Google Research	United States	1,500,000,000	bee-mouse	Speculative	unconfirmed
BLIP-2 (Q-Former)	2023-01-30	Salesforce Research	United States	1,480,000,000	bee-mouse	Confident	–
LDM-1.45B	2021-12-20	Heidelberg University,Runway	Germany; United States	1,450,000,000	bee-mouse	Confident	–
Stable Diffusion (LDM-KL-8-G)	2022-04-13	Runway,Ludwig Maximilian University of Munich,Heid	Germany; United States	1,450,000,000	bee-mouse	Confident	–
SEER	2021-07-29	Facebook AI Research,INRIA	France; United States	1,300,000,000	bee-mouse	Confident	–
InstructGPT 1.3B	2022-01-27	OpenAI	United States	1,300,000,000	bee-mouse	Confident	–
Megatron-LM (1.2B)	2019-09-17	Nvidia	United States	1,200,000,000	bee-mouse	Confident	–
RDT-1B	2024-10-10	Tsinghua University	China	1,200,000,000	bee-mouse	Likely	–
Gato	2022-05-12	DeepMind	United Kingdom	1,180,000,000	bee-mouse	Confident	–
DINOv2	2023-04-14	Facebook AI Research,INRIA	France; United States	1,140,000,000	bee-mouse	Confident	–
Diffusion Renderer	2025-03-22	Nvidia,University of Toronto,Vector Institute,Univ	Canada; United States	1,100,000,000	bee-mouse	Confident	–
InternImage	2022-11-10	Shanghai AI Lab,Tsinghua University,Nanjing Univer	China; Hong Kong	1,080,000,000	bee-mouse	Confident	–
EVA-01	2022-11-14	Beijing Academy of Artificial Intelligence / BAAI,	China	1,011,000,000	bee-mouse	Confident	–
Unsupervised High-level Feature Learner	2012-07-12	Google	United States	1,000,000,000	bee-mouse	Likely	sparse (MoE)
Conformer + Wav2vec 2.0 + Noisy Student	2020-10-20	Google,Google Research,Google Brain	United States	1,000,000,000	bee-mouse	Confident	–
HuBERT	2021-07-27	Facebook AI Research	France; United States	1,000,000,000	bee-mouse	Speculative	unconfirmed
W2v-BERT	2021-08-07	Google Brain,Massachusetts Institute of Technology	United States	1,000,000,000	bee-mouse	Confident	–
AudioGen	2023-03-05	Meta AI,Hebrew University of Jerusalem	Israel; United States	1,000,000,000	bee-mouse	Likely	–
VideoMAE V2	2023-03-29	Nanjing University,Shenzhen Institute of Advanced	China	1,000,000,000	bee-mouse	Confident	–
TransE	2013-12-05	Universite de Technologie de Compiègne – CNRS,Goog	France; United States	942,000,000	bee-mouse	Speculative	unconfirmed
ImageBind	2023-05-09	Meta AI	United States	932,000,000	bee-mouse	Likely	–
Big Transfer (BiT-L)	2019-12-24	Google Brain	United States	928,000,000	bee-mouse	–	–
Florence	2021-11-22	Microsoft	United States	893,000,000	bee-mouse	Confident	–
NÜWA	2021-11-24	Microsoft Research,Peking University	China; United States	870,000,000	bee-mouse	Confident	–
MusicLM	2023-01-26	Google	United States	860,000,000	bee-mouse	Confident	–
ResNeXt-101 32x48d	2018-05-02	Facebook	United States	829,000,000	bee-mouse	Confident	–
FixRes ResNeXt-101 WSL	2019-06-14	Facebook AI	United States	829,000,000	bee-mouse	–	–
ALIGN	2021-06-11	Google Research	United States	820,000,000	bee-mouse	Confident	–
Statement Curriculum Learning	2022-03-02	OpenAI	United States	774,000,000	bee-mouse	Likely	–
CodeT5-large	2022-07-05	Salesforce	United States	770,000,000	bee-mouse	Likely	–
DITTO	2022-06-06	Tsinghua University,Apple,Westlake University,Chin	China; Hong Kong; United States	750,000,000	bee-mouse	Confident	–
FGN	2025-06-12	Google DeepMind	United States	720,000,000	bee-mouse	Confident	–
data2vec (vision)	2022-01-20	Meta AI	United States	705,134,592	bee-mouse	–	–
data2vec (speech)	2022-01-20	Meta AI	United States	705,134,592	bee-mouse	–	–
data2vec (language)	2022-01-20	Meta AI	United States	705,134,592	bee-mouse	Likely	–
Tranception	2022-05-27	University of Oxford,Harvard Medical School,Cohere	Canada; United Kingdom; United States	700,000,000	bee-mouse	Confident	–
Word2Vec (large)	2013-10-16	Google	United States	692,000,000	bee-mouse	Likely	–
DiT-XL/2	2023-03-02	New York University (NYU),University of California	United States	675,000,000	bee-mouse	Confident	–
DiT-XL/2 + CADS	2023-10-26	ETH Zurich,Disney Research	Switzerland; United States	675,000,000	bee-mouse	Likely	–
OmegaPLM	2022-07-22	Massachusetts Institute of Technology (MIT),Westla	China; United States	670,000,000	bee-mouse	Confident	–
XLM	2019-06-01	Facebook	United States	665,000,000	bee-mouse	–	–
ESM1b	2020-12-15	Facebook AI Research,New York University (NYU)	France; United States	652,400,000	bee-mouse	Confident	–
Segment Anything Model	2023-04-05	Meta AI	United States	636,000,000	bee-mouse	Confident	–
ViT-Huge/14	2020-10-22	Google Brain,Google Research	United States	632,000,000	bee-mouse	Confident	–
Masked Autoencoders ViT-H	2021-11-11	Facebook AI Research	France; United States	632,000,000	bee-mouse	Speculative	unconfirmed
Retrieval-Augmented Generator	2020-05-22	Facebook,New York University (NYU),University Coll	United Kingdom; United States	626,000,000	bee-mouse	Confident	–
DLDL (PASCAL)	2016-11-06	University of Oxford	United Kingdom	564,000,000	bee-mouse	Likely	–
ADM	2021-05-11	OpenAI	United States	559,000,000	bee-mouse	Confident	–
TrOCR	2021-09-21	Beihang University,Microsoft Research Asia	China	558,000,000	bee-mouse	Confident	–
XLM-RoBERTa	2019-11-05	Facebook AI	United States	550,000,000	bee-mouse	Confident	–
LUKE	2020-10-02	University of Washington,National Institute of Inf	Japan; United States	483,000,000	bee-mouse	Likely	–
EfficientNet-L2	2019-05-28	Google	United States	480,000,000	bee-mouse	–	–
Noisy Student (L2)	2019-11-11	Carnegie Mellon University (CMU),Google	United States	480,000,000	bee-mouse	Confident	–
Meta Pseudo Labels	2021-03-01	Google Brain,Google AI	United States	480,000,000	bee-mouse	Likely	–
EMDR	2021-06-09	Mila – Quebec AI (originally Montreal Institute fo	Canada; United Kingdom	440,000,000	bee-mouse	Confident	–
NMT Transformer 437M	2019-02-28	Google,Bar-Ilan University	Israel; United States	437,700,000	bee-mouse	Confident	–
Adaptive Inputs + LayerDrop	2019-09-25	Facebook AI Research,LORIA	France; United States	423,000,000	bee-mouse	Confident	–
ProtBERT-BFD	2021-05-04	Technical University of Munich,Nvidia,Seoul Nation	China; Germany; South Korea; United Stat	420,000,000	bee-mouse	Confident	–
OntoProtein	2022-01-23	Zhejiang University (ZJU)	China	420,000,000	bee-mouse	Likely	–
DeBERTaV3large + KEAR	2022-05-04	Microsoft	United States	418,000,000	bee-mouse	Confident	–
BART-large	2019-10-29	Facebook AI	United States	406,291,456	bee-mouse	Likely	–
SigLIP 400M	2023-03-27	Google DeepMind	United States	400,000,000	bee-mouse	Confident	–
BERT-Large-CAS (PTB+WT2+WT103)	2019-04-20	Amazon	United States	395,000,000	bee-mouse	Likely	–
MoCo	2019-11-13	Facebook AI	United States	375,000,000	fly-bee	–	–
SimCLR	2020-02-13	Google Brain	United States	375,000,000	fly-bee	–	–
CLIP (ViT L/14@336px)	2021-01-05	OpenAI	United States	370,000,000	fly-bee	Confident	–
RoBERTa Large	2019-07-01	Facebook,University of Washington	United States	355,000,000	fly-bee	Confident	–
AbLang (heavy sequences)	2022-01-22	University of Oxford	United Kingdom	355,000,000	fly-bee	Confident	–
GPT-2 Medium (FlashAttention)	2022-05-27	Stanford University,University at Buffalo	United States	355,000,000	fly-bee	Confident	–
VALL-E	2023-01-05	Microsoft	United States	353,000,000	fly-bee	Speculative	unconfirmed
BERT-Large	2018-10-11	Google	United States	340,000,000	fly-bee	Confident	–
XLNet	2019-06-01	Carnegie Mellon University (CMU),Google Brain	United States	340,000,000	fly-bee	Confident	–
ERNIE-GEN (large)	2020-08-06	Baidu	China	340,000,000	fly-bee	Speculative	unconfirmed
CamemBERT	2019-11-10	Facebook,INRIA,Sorbonne University	France; United States	335,000,000	fly-bee	Confident	–
ELECTRA	2020-03-23	Stanford University,Google,Google Brain	United States	335,000,000	fly-bee	Likely	–
German ELECTRA Large	2020-10-21	deepset,Bayerische Staatsbibliothek Muenchen	Germany	335,000,000	fly-bee	Confident	–
ALM 1.0	2022-11-28	Beijing Academy of Artificial Intelligence / BAAI	China	335,000,000	fly-bee	Speculative	unconfirmed
MT-DNN	2019-01-31	Microsoft	United States	330,000,000	fly-bee	Confident	–
Fusion in Encoder	2022-11-18	Samsung	South Korea	330,000,000	fly-bee	Likely	–
wave2vec 2.0 LARGE	2020-10-22	Facebook	United States	317,000,000	fly-bee	Confident	–
CPC v2	2019-05-22	DeepMind,University of California (UC) Berkeley	United Kingdom; United States	303,000,000	fly-bee	–	–
Transformer – LibriVox + Decoding/Rescoring	2019-11-19	Facebook	United States	296,000,000	fly-bee	Confident	–
DDPM-IP (CelebA)	2023-01-27	Utrecht University	Netherlands	295,000,000	fly-bee	Likely	–
GNMT	2016-09-26	Google	United States	278,000,000	fly-bee	Likely	sparse (MoE)
base LM+GNN+kNN	2021-10-17	Shannon.AI,Nanjing University,Nanyang Technologica	China; Singapore	274,000,000	fly-bee	Confident	–
YOLO	2015-06-08	University of Washington,Allen Institute for AI,Fa	France; United States	271,684,800	fly-bee	–	–
Transformer-XL (257M)	2019-01-09	Carnegie Mellon University (CMU),Google Brain	United States	257,000,000	fly-bee	Speculative	unconfirmed
Transformer-XL + RMS dynamic eval	2019-04-17	University of Edinburgh	United Kingdom	257,000,000	fly-bee	Confident	–
Transformer-XL Large + Phrase Induction	2019-06-04	Massachusetts Institute of Technology (MIT),Univer	United States	257,000,000	fly-bee	Speculative	unconfirmed
Segatron-XL large, M=384 + HCP	2022-03-21	Microsoft Research,University of Waterloo	Canada; United States	257,000,000	fly-bee	Confident	–
Denoising Diffusion Probabilistic Models (LSUN Bedroom)	2021-06-11	University of California (UC) Berkeley	United States	256,000,000	fly-bee	Confident	–
Pangu-Weather	2023-07-05	Huawei	China	256,000,000	fly-bee	Confident	–
SPN (ImageNet 128)	2018-12-04	Google Brain,DeepMind	United Kingdom; United States	250,000,000	fly-bee	Confident	–
S4	2021-10-31	Stanford University	United States	249,000,000	fly-bee	Likely	–
Base LM + kNN LM + Continuous Cache	2019-11-01	Stanford University,Facebook AI Research	France; United States	247,000,000	fly-bee	Likely	–
ERNIE-Doc (247M)	2020-12-31	Baidu	China	247,000,000	fly-bee	Confident	–
Adaptive Input Transformer + RD	2021-06-28	Microsoft Research Asia,Soochow University	China; Taiwan	247,000,000	fly-bee	Likely	–
Transformer (Adaptive Input Embeddings) WT103	2018-09-28	Facebook AI Research	France; United States	247,000,000	fly-bee	Confident	–
TaLK Convolution	2020-02-08	Carleton University	Canada	240,000,000	fly-bee	Likely	–
ALBERT-xxlarge	2020-02-09	Toyota Technological Institute at Chicago,Google	United States	235,000,000	fly-bee	Speculative	unconfirmed
SRU++ Large	2021-02-24	ASAPP	United States	234,000,000	fly-bee	Confident	–
GR-2	2024-10-08	ByteDance	China	230,000,000	fly-bee	Confident	–
MaskGIT (ImageNet)	2022-02-08	Google Research	United States	227,000,000	fly-bee	Confident	–
MetNet	2020-03-24	Google	United States	225,000,000	fly-bee	Speculative	unconfirmed
FFN SwiGLU	2020-02-14	Google	United States	220,000,000	fly-bee	Confident	–
CodeT5-base	2021-11-01	Salesforce,Nanyang Technological University	Singapore; United States	220,000,000	fly-bee	Likely	–
Transformer	2017-06-12	Google Research,Google Brain	United States	213,000,000	fly-bee	Confident	–
Sandwich Transformer	2019-11-10	Allen Institute for AI,Facebook AI Research	France; United States	209,000,000	fly-bee	Speculative	unconfirmed
EfficientNetV2-XL	2021-06-23	Google,Google Brain	United States	208,000,000	fly-bee	Confident	–
Word2Vec (small)	2013-10-16	Google	United States	207,600,000	fly-bee	–	–
Multilingual DNN	2013-05-26	Google	United States	206,899,200	fly-bee	Confident	–
DreamerV3	2023-01-10	DeepMind,University of Toronto	Canada; United Kingdom	200,000,000	fly-bee	Likely	–
DL scaling speech	2017-12-01	Baidu	China	193,000,000	fly-bee	Likely	–
ResNeXt-101 Billion-scale	2019-05-02	Facebook AI	United States	193,000,000	fly-bee	–	–
(ensemble): AWD-LSTM-DOC (fin) × 5 (WT2)	2018-08-30	NTT Communication Science Laboratories,Tohoku Univ	Japan	185,000,000	fly-bee	Confident	–
TrellisNet	2018-10-15	Carnegie Mellon University (CMU),Bosch Center for	Germany; United States	180,000,000	fly-bee	Confident	–
DL scaling LM	2017-12-01	Baidu	China	177,000,000	fly-bee	Speculative	unconfirmed
EXAONE Path 2.0	2025-07-09	LG AI Research	South Korea	175,000,000	fly-bee	Confident	–
OpenAI Five Rerun	2019-12-13	OpenAI	United States	159,000,000	fly-bee	Confident	–
OpenAI Five	2019-12-13	OpenAI	United States	159,000,000	fly-bee	Confident	–
4 layer QRNN (h=2500)	2018-03-22	Salesforce Research	United States	151,000,000	fly-bee	Likely	–
TransformerXL + spectrum control	2020-03-11	University of California Los Angeles (UCLA),JD.com	China; United States	151,000,000	fly-bee	Speculative	unconfirmed
AlphaGeometry	2024-01-17	Google DeepMind,New York University (NYU)	United States	151,000,000	fly-bee	Confident	–
PermuteFormer	2021-09-06	Peking University	China	149,697,024	fly-bee	Speculative	unconfirmed
Fragment embedding	2014-06-21	Stanford University	United States	144,496,000	fly-bee	Likely	–
OverFeat	2013-12-21	New York University (NYU)	United States	144,000,000	fly-bee	Confident	–
VGG19	2014-09-04	University of Oxford	United Kingdom	144,000,000	fly-bee	Likely	–
LRCN	2014-11-07	UT Austin,University of Massachusetts Lowell,Unive	United States	142,552,000	fly-bee	–	–
Transformer-XL DeFINE (141M)	2019-11-27	University of Washington,Allen Institute for AI	United States	141,000,000	fly-bee	Speculative	unconfirmed
Deep Autoencoders	2011-04-29	University of Toronto	Canada	139,808,256	fly-bee	Confident	–
AlphaStar	2019-10-30	DeepMind	United Kingdom	139,000,000	fly-bee	Confident	–
VGG16	2014-09-04	University of Oxford	United Kingdom	138,000,000	fly-bee	Confident	–
VGG-Face	2015-01-01	University of Oxford	United Kingdom	138,000,000	fly-bee	Confident	–
SAF R-CNN	2015-10-28	Beijing Institute of Technology,Sun Yat-sen Univer	China; Singapore	138,000,000	fly-bee	Likely	–
Template Adaptation	2016-03-12	University of Oxford	United Kingdom	138,000,000	fly-bee	Likely	–
LRR-4X	2016-05-08	UC Irvine	United States	138,000,000	fly-bee	Likely	–
CMS-RCNN	2016-06-17	IEEE	Multinational	138,000,000	fly-bee	Likely	–
QRNN	2018-02-01	Salesforce Research	United States	135,000,000	fly-bee	Likely	–
Multiresolution CNN	2014-06-23	Google,Stanford University	United States	126,125,568	fly-bee	–	–
FTW (For The Win)	2018-07-03	DeepMind	United Kingdom	126,001,330	fly-bee	Speculative	unconfirmed
Feedback Transformer	2020-02-21	LORIA,University of Lorraine,Facebook AI Research	France; United States	126,000,000	fly-bee	Speculative	unconfirmed
KEPLER	2020-11-23	Tsinghua University,Mila – Quebec AI (originally M	Canada; China; France; United States	125,000,000	fly-bee	Confident	–
DL scaling Image	2017-12-01	Baidu	China	121,000,000	fly-bee	Confident	–
GloVe (32B)	2014-01-01	Stanford University	United States	120,000,000	fly-bee	–	–
GloVe (6B)	2014-01-01	Stanford University	United States	120,000,000	fly-bee	–	–
Conformer	2020-05-16	Google	United States	118,800,000	fly-bee	Confident	–
GPT-1	2018-06-01	OpenAI	United States	117,000,000	fly-bee	Likely	–
ContextNet	2020-05-07	Google	United States	112,700,000	fly-bee	Likely	–
SciBERT	2019-03-26	Allen Institute for AI	United States	110,000,000	fly-bee	Confident	–
DNABERT	2021-08-15	Northeastern University	United States	110,000,000	fly-bee	Confident	–
Contriever	2021-12-16	Meta AI,University College London (UCL),PSL Univer	France; United Kingdom; United States	110,000,000	fly-bee	Likely	–
BERT-RBP	2022-04-07	Waseda University	Japan	110,000,000	fly-bee	Confident	–
RoFormer	2023-11-08	Zhuiyi Technology	China	110,000,000	fly-bee	Confident	–
LSTM LM	2012-09-09	RWTH Aachen University	Germany	102,720,000	fly-bee	Speculative	unconfirmed
ReLU-Speech	2013-05-26	Google,University of Toronto,New York University (	Canada; United States	101,706,240	fly-bee	Likely	–
STORM-B/8	2024-12-31	University of Southern California,Georgia Institut	United States	100,598,707	fly-bee	Confident	–
Support Vector Machines	1995-09-01	AT&T,Bell Laboratories	United States	100,000,000	fly-bee	–	not a neural network
GPU DBNs	2009-06-15	Stanford University	United States	100,000,000	fly-bee	Confident	–
MSA Transformer	2021-02-13	Facebook AI Research,University of California (UC)	France; United States	100,000,000	fly-bee	Likely	–
YOLOX-X	2021-08-06	Megvii Inc	China	99,100,000	fly-bee	Likely	–
DeLighT	2020-08-03	University of Washington,Allen Institute for AI,Fa	France; United States	99,000,000	fly-bee	Likely	–
ELMo	2018-02-01	University of Washington,Allen Institute for AI	United States	94,000,000	fly-bee	Speculative	unconfirmed
AlphaFold 2	2020-11-30	DeepMind	United Kingdom	93,000,000	fly-bee	Likely	–
AlphaMissense	2023-09-22	Google DeepMind	United States	93,000,000	fly-bee	Likely	–
Octo-Base	2024-05-20	University of California (UC) Berkeley,Stanford Un	United Kingdom; United States	93,000,000	fly-bee	Confident	–
PolyNet	2016-11-17	Chinese University of Hong Kong (CUHK)	Hong Kong	92,000,000	fly-bee	Likely	–
Transformer local-attention (NesT-B)	2021-05-26	Google Cloud,Google Research	United States	90,100,000	fly-bee	Confident	–
Transformer + Simple Recurrent Unit	2018-09-17	ASAPP,Cornell University,Google,Princeton Universi	United States	90,000,000	fly-bee	Confident	–
NASNet-A	2017-07-21	Google Brain	United States	89,000,000	fly-bee	–	–
CLIP (ResNet-50)	2021-01-05	OpenAI	United States	88,600,000	fly-bee	Likely	–
Detic	2022-01-07	Meta AI,University of Texas at Austin	United States	88,000,000	fly-bee	Speculative	unconfirmed
MSRA (C, PReLU)	2015-02-06	Microsoft Research	United States	87,048,800	fly-bee	Confident	–
BigBiGAN	2019-07-04	Google	United States	86,000,000	fly-bee	–	–
ViT-Base/32	2020-10-22	Google Brain	United States	86,000,000	fly-bee	–	–
DeiT-B	2021-01-15	Meta AI,Sorbonne University	France; United States	86,000,000	fly-bee	Confident	–
CaLM	2022-12-19	University of Oxford	United Kingdom	86,000,000	fly-bee	Likely	–
Tensor-Transformer(1core)+PN (WT103)	2020-03-17	University of California (UC) Berkeley	United States	85,300,000	fly-bee	Confident	–
ResNeXt-101 (64×4d)	2016-11-16	University of California San Diego,Facebook	United States	83,000,000	fly-bee	Confident	–
RBM Image Classifier	2009-04-08	University of Toronto	Canada	80,000,000	fly-bee	Likely	–
Routing Transformer (WT-103)	2020-03-12	Google Research	United States	79,500,000	fly-bee	Confident	–
Big-Little Net	2018-07-10	IBM	United States	77,360,000	fly-bee	Likely	–
EfficientDet	2020-07-27	Google Brain	United States	77,000,000	fly-bee	–	–
CODEFUSION (Python)	2023-10-26	Microsoft,Microsoft Research	United States	75,000,000	fly-bee	Confident	–
RNN LM	2010-09-26	Johns Hopkins University	United States	70,265,000	fly-bee	Speculative	unconfirmed
R-CNN (T-net)	2013-11-11	University of California (UC) Berkeley	United States	69,003,872	fly-bee	–	–
Variational (untied weights, MC) LSTM (Large)	2015-12-16	University of Cambridge	United Kingdom	66,000,000	fly-bee	Confident	–
DistilBERT	2019-10-02	Hugging Face	United States	66,000,000	fly-bee	Confident	–
Theseus 6/768	2020-02-07	University of California San Diego,Beihang Univers	China; United States	66,000,000	fly-bee	Confident	–
ResNet-152 (ImageNet)	2015-12-10	Microsoft	United States	60,200,000	fly-bee	Confident	–
AlexNet	2012-09-30	University of Toronto	Canada	60,000,000	fly-bee	Confident	–
DETR	2020-05-26	Facebook	United States	60,000,000	fly-bee	Confident	–
YOLOv3	2018-04-08	University of Washington	United States	56,933,216	fly-bee	Likely	–
Inception-ResNet-V2	2016-02-23	Google	United States	56,000,000	fly-bee	–	–
Transformer ELMo	2019-01-01	Allen Institute for AI,University of Washington	United States	56,000,000	fly-bee	Confident	–
NAS with base 8 and shared embeddings	2016-11-05	Google Brain	United States	54,000,000	< fly	Confident	–
RetinaNet-R101	2017-08-07	Facebook AI Research	France; United States	53,000,000	< fly	Confident	–
AWD-LSTM+WT+Cache+IOG (WT2)	2017-09-26	NTT Communication Science Laboratories	Japan	53,000,000	< fly	Confident	–
Hierarchical Scene Labeling (Stanford Background)	2013-08-01	New York University (NYU)	United States	51,609,600	< fly	Confident	–
VD-LSTM+REAL Large	2016-11-04	Salesforce Research,Stanford University	United States	51,000,000	< fly	Confident	–
YOLOv2	2016-12-25	University of Washington,Allen Institute for AI	United States	51,000,000	< fly	–	–
Dropout-LSTM+Noise(Bernoulli) (WT2)	2018-05-03	Columbia University,New York University (NYU),Prin	United States	51,000,000	< fly	Likely	–
LSTM + dynamic eval	2017-09-21	University of Edinburgh	United Kingdom	50,000,000	< fly	Likely	–
Dropout (TIMIT)	2012-06-03	University of Toronto	Canada	48,840,185	< fly	–	–
Dropout: SVHN	2014-06-01	University of Toronto	Canada	47,795,232	< fly	Confident	–
DistBelief Speech	2012-12-03	Google	United States	47,185,920	< fly	Speculative	unconfirmed
AlphaGo Zero	2017-10-18	DeepMind	United Kingdom	46,400,244	< fly	Confident	–
Fine-tuned-AWD-LSTM-DOC (fin)	2018-11-12	Samsung R&D Institute Russia	Russia	46,000,000	< fly	Confident	–
CT-MoS (WT2)	2020-12-25	Google,National Tsing Hua University	Taiwan; United States	45,000,000	< fly	Confident	–
JFT	2017-07-10	Google Research,Carnegie Mellon University (CMU)	United States	44,654,504	< fly	Confident	–
ResNet-101 (ImageNet)	2015-12-10	Microsoft	United States	44,500,000	< fly	Confident	–
HR-ResNet101	2016-12-13	Carnegie Mellon University (CMU)	United States	44,500,000	< fly	Confident	–
3DMM-CNN	2016-12-15	University of Southern California	United States	44,500,000	< fly	Confident	–
LaNet-L (CIFAR-10)	2019-06-17	Brown University,Facebook	United States	44,100,000	< fly	Confident	–
ProBERTa	2020-09-01	University of Illinois Urbana-Champaign (UIUC),Ree	United States	44,000,000	< fly	Confident	–
PreTrans-3L-250H	2013-03-22	University of Toronto	Canada	43,000,000	< fly	–	–
Inceptionv4	2016-02-23	Google	United States	43,000,000	< fly	–	–
GL-LWGC-AWD-MoS-LSTM + dynamic evaluation (WT2)	2017-08-29	Ben-Gurion University of the Negev	Israel	38,000,000	< fly	Confident	–
U-Net	2015-05-18	University of Freiburg	Germany	37,676,160	< fly	Confident	–
NASv3 (CIFAR-10)	2016-11-05	Google Brain	United States	37,400,000	< fly	Likely	–
MuZero	2019-11-19	DeepMind	United Kingdom	36,864,000	< fly	Speculative	unconfirmed
AWD-LSTM-MoS + dynamic evaluation (WT2, 2017)	2017-11-10	Carnegie Mellon University (CMU)	United States	35,000,000	< fly	Likely	–
AWD-LSTM-MoS+PDR + dynamic evaluation (WT2)	2018-08-14	IBM	United States	35,000,000	< fly	Confident	–
AWD-LSTM-MoS + dynamic evaluation (WT2, 2018)	2018-09-18	Peking University,Microsoft Research Asia	China	35,000,000	< fly	Confident	–
AWD-LSTM + MoS + Partial Shuffled	2019-06-10	University of Texas at Austin	United States	35,000,000	< fly	Confident	–
Mogrifier (d2, MoS2, MC) + dynamic eval	2019-09-04	DeepMind,University of Oxford	United Kingdom	35,000,000	< fly	Confident	–
Mogrifier RLSTM (WT2)	2022-11-03	DeepMind	United Kingdom	35,000,000	< fly	Confident	–
RT-1	2022-12-13	Google	United States	35,000,000	< fly	Confident	–
DenseNet-264	2016-08-25	Tsinghua University,Facebook AI Research,Cornell U	China; France; United States	34,000,000	< fly	Confident	–
RetinaNet-R50	2017-08-07	Facebook AI Research	France; United States	34,000,000	< fly	–	–
Fraternal dropout + AWD-LSTM 3-layer (WT2)	2017-10-31	Jagiellonian University,Mila – Quebec AI (original	Canada; Poland	34,000,000	< fly	Likely	–
AWD-LSTM-DRILL + dynamic evaluation† (WT2)	2019-05-14	IDIAP	Switzerland	34,000,000	< fly	Likely	–
AWD-LSTM – 3-layer LSTM (tied) + continuous cache pointer (WT2)	2017-08-07	Salesforce Research	United States	33,000,000	< fly	Confident	–
TCAN (WT2)	2020-02-28	Nanjing University,Ant Group	China	33,000,000	< fly	Confident	–
MMLSTM (WT-2)	2019-12-05	Beijing University of Posts and Telecommunications	China; United Kingdom	32,300,000	< fly	Likely	–
Paragraph Vector	2014-05-14	Google	United States	32,000,000	< fly	Confident	–
aLSTM(depth-2)+RecurrentPolicy (WT2)	2018-05-22	University of Manchester,Alan Turing Institute	United Kingdom	32,000,000	< fly	Confident	–
StyleGAN2	2019-12-03	Nvidia,Aalto University	Finland; United States	30,000,000	< fly	Confident	–
UDSMProt	2019-09-04	Fraunhofer Heinrich Hertz Institute	Germany	28,303,800	< fly	Likely	–
SENet (ImageNet)	2017-09-05	Chinese Academy of Sciences,University of Oxford	China; United Kingdom	28,100,000	< fly	–	–
Perceiver IO (optical flow)	2020-02-08	DeepMind	United Kingdom	27,900,000	< fly	Confident	–
Fractional Max-Pooling	2014-12-18	University of Warwick	United Kingdom	27,000,000	< fly	Likely	–
Hierarchical Reasoning Model (HPM)	2025-08-04	Sapient Intelligence	Singapore	27,000,000	< fly	Confident	–
RNN-WER	2014-06-22	DeepMind,University of Toronto	Canada; United Kingdom	26,500,000	< fly	Likely	–
StyleGAN	2018-12-12	Nvidia	United States	26,200,000	< fly	Confident	–
PyramidNet	2017-09-06	Korea Advanced Institute of Science and Technology	South Korea	26,000,000	< fly	Likely	–
ResNeXt-50	2016-11-16	University of California San Diego,Facebook	United States	25,000,000	< fly	Likely	–
AWD-LSTM	2017-07-18	DeepMind,University of Oxford	United Kingdom	24,000,000	< fly	Confident	–
Inception v3	2015-12-02	Google,University College London (UCL)	United Kingdom; United States	23,626,728	< fly	Likely	–
Mamba-24M (SC09)	2023-12-01	Carnegie Mellon University (CMU),Princeton Univers	United States	23,400,000	< fly	Confident	–
WeNet (Penn Treebank)	2019-04-08	Amazon	United States	23,000,000	< fly	Likely	–
EN^2AS with performance reward	2019-07-22	Beijing Institute of Technology,University of Tech	Australia; China	23,000,000	< fly	–	–
NAS+ESS (23M)	2020-05-06	Northeastern University (China),NiuTrans Research,	China	23,000,000	< fly	Confident	–
Xception	2016-10-07	Google	United States	22,855,952	< fly	Confident	–
MetaMimic	2018-10-11	Google	United States	22,000,000	< fly	–	–
MMLSTM (PTB)	2019-12-05	Beijing University of Posts and Telecommunications	China; United Kingdom	21,300,000	< fly	Likely	–
Pointer Sentinel-LSTM (medium)	2016-09-26	MetaMind Inc,Salesforce	United States	21,000,000	< fly	Confident	–
DiffDock	2022-10-04	Massachusetts Institute of Technology (MIT)	United States	20,240,000	< fly	Likely	–
EI-REHN-1000D	2017-08-14	Korea Advanced Institute of Science and Technology	South Korea	19,000,000	< fly	Confident	–
TC-DNN-BLSTM-DNN	2015-04-06	Carnegie Mellon University (CMU)	United States	18,413,568	< fly	Confident	–
OR-WideResNet	2017-01-07	Duke University,University of Chinese Academy of S	China; United States	18,200,000	< fly	Confident	–
BP-DBN	2009-01-01	University of Toronto	Canada	18,030,592	< fly	Confident	–
ALBERT	2019-09-26	Toyota Technological Institute at Chicago,Google R	United States	18,000,000	< fly	Confident	–
MedBERT	2021-05-20	Peng Cheng Laboratory,University of Texas at Houst	China; United States	17,000,000	< fly	Likely	–
LCNP NORB	2009-11-22			16,818,176	< fly	Confident	–
LF-MMI	2016-09-08	Johns Hopkins University,Cornell University	United States	16,600,000	< fly	Confident	–
AlphaFold	2020-01-15	DeepMind	United Kingdom	16,340,840	< fly	Speculative	unconfirmed
GNoME for crystal discovery	2023-11-29	Google DeepMind	United States	16,240,000	< fly	Likely	–
ReLU (NORB)	2010-06-15	University of Toronto	Canada	16,210,006	< fly	–	–
ProteinBERT	2022-02-10	Hebrew University of Jerusalem,Ben-Gurion Universi	Israel; United States	16,000,000	< fly	Confident	–
Eve	2021-10-27	Harvard Medical School,University of Oxford	United Kingdom; United States	15,010,300	< fly	Likely	–
LCNP LabelMe	2009-11-22	University of Bonn	Germany	13,729,792	< fly	Confident	–
BatchNorm	2015-06-15	Google	United States	13,600,000	< fly	Confident	–
Fold2Seq	2021-06-24	IBM,Texas A&M	United States	12,427,904	< fly	Speculative	unconfirmed
NPLM (AP News)	2003-03-15	University of Montreal / Université de Montréal	Canada	11,904,264	< fly	Confident	–
LCNP MNIST	2009-11-22			11,616,256	< fly	Confident	–
ISS	2017-09-15	Duke University,Microsoft	United States	11,100,000	< fly	Confident	–
ResNet-1001	2016-09-17	Microsoft	United States	10,200,000	< fly	–	–
RNN+LDA+KN5+cache	2012-12-01	Microsoft,Brno University of Technology	Czechia; United States	9,000,000	< fly	–	–
Gated HORNN (3rd order)	2016-04-30	York University	Canada	8,970,000	< fly	–	–
AlphaGo Fan	2015-10-01	DeepMind	United Kingdom	8,209,984	< fly	Likely	–
genCNN + dyn eval	2015-03-17	Chinese Academy of Sciences,Huawei Noah’s Ark Lab,	China; Ireland	8,000,000	< fly	Speculative	unconfirmed
Once for All	2020-04-29	MIT-IBM Watson AI Lab,Massachusetts Institute of T	United States	7,700,000	< fly	Confident	–
Multi-cell LSTM	2018-11-15	University of Hyderabad	India	7,200,000	< fly	Likely	–
Feedforward NN	2010-05-13	University of Montreal / Université de Montréal	Canada	7,082,000	< fly	Confident	–
True-Regularization+Finetune+Dynamic-Eval	2019-04-08	Mobvoi,Williams College	China; United States	7,000,000	< fly	Confident	–
Neural LM	2000-11-28	University of Montreal / Université de Montréal	Canada	6,906,980	< fly	Confident	–
GoogLeNet / InceptionV1	2014-09-17	Google,University of Michigan,University of North	United States	6,797,700	< fly	Confident	–
HyenaDNA	2023-06-27	Stanford University,Harvard University,Mila – Queb	Canada; United States	6,600,000	< fly	Confident	–
Dropout (MNIST)	2012-06-03	University of Toronto	Canada	5,594,010	< fly	Confident	–
SimpleNet	2016-08-22	Sensifai,Islamic Azad University,Technicolor R&I,I	Belgium; France; Iran (Islamic Republic	5,480,000	< fly	Confident	–
AlphaX-1	2019-10-02	Facebook AI Research,Brown University	France; United States	5,400,000	< fly	Confident	–
3DDFA	2015-11-23	Chinese Academy of Sciences,Michigan State Univers	China; United States	5,355,584	< fly	Confident	–
Deep CNN + COTS	2015-07-26	IEEE	Multinational	5,006,000	< fly	Confident	–
DCNN	2015-08-07	University of Maryland,Rutgers University	United States	5,006,000	< fly	Confident	–
NLP from scratch	2011-11-08	NEC Laboratories,Princeton University	United States	5,000,000	< fly	–	–
SPN-4+KN5	2014-09-14	Singapore University of Technology & Design,DSO Na	Singapore	5,000,000	< fly	Likely	–
6-Act Tether	2021-08-03	Facebook AI Research,Georgia Institute of Technolo	France; United States	5,000,000	< fly	Confident	–
High Performance CNN (NORB)	2011-07-16	IDSIA,SUPSI	Switzerland	4,878,300	< fly	Likely	–
MobileNet	2017-04-17	Google	United States	4,200,000	< fly	–	–
NPLM (Brown)	2003-03-15	University of Montreal / Université de Montréal	Canada	4,124,233	< fly	Confident	–
LMICA	2004-12-01			4,096,000	< fly	Confident	–
PoE MNIST	2000-11-28	University College London (UCL)	United Kingdom	3,925,310	< fly	Confident	–
Dimensionality Reduction	2006-07-18	University of Toronto	Canada	3,800,000	< fly	–	–
GOAT	2021-07-27	DeepMind	United Kingdom	3,472,816	< fly	Speculative	unconfirmed
MobileNetV2	2018-06-18	Google	United States	3,400,000	< fly	–	–
Big-Little Net (speech)	2018-07-10	IBM	United States	3,320,000	< fly	Speculative	unconfirmed
DLWP	2021-02-09	University of Washington,Microsoft Research	United States	2,676,376	< fly	Confident	–
BIDAF	2016-11-05	University of Washington,Allen Institute for AI	United States	2,600,000	< fly	Confident	–
DeepStack	2017-01-06	University of Alberta,Charles University,Czech Tec	Canada; Czechia	2,500,000	< fly	Speculative	unconfirmed
KataGo	2019-02-27	Jane Street	United States	2,500,000	< fly	Speculative	unconfirmed
ShuffleNet v1	2017-07-03	Megvii Inc	China	2,430,000	< fly	–	–
ADAM (CIFAR-10)	2014-12-22	University of Amsterdam,OpenAI,University of Toron	Canada; Netherlands; United States	2,370,000	< fly	Confident	–
StyleGAN3-T	2021-06-21	Nvidia,Aalto University	Finland; United States	2,230,000	< fly	Confident	–
LSTM+NeuralCache	2018-09-24	KU Leuven,ESAT – PSI,Apple	Belgium; United States	2,100,000	< fly	Likely	–
HLBL	2008-12-08	University of Toronto	Canada	1,846,400	< fly	Confident	–
ResNet-110 (CIFAR-10)	2015-12-10	Microsoft	United States	1,700,000	< fly	–	–
Dueling DQN	2016-04-05	Google DeepMind	United States	1,700,000	< fly	Confident	–
DQN-2015	2015-02-25	Google	United States	1,693,362	< fly	–	–
Rational DQN Average	2021-02-18	TU Darmstadt	Germany	1,683,456	< fly	–	–
IBM-5	1993-06-15	IBM	United States	1,658,364	< fly	–	–
Deep Belief Nets	2006-07-18	University of Toronto,National University of Singa	Canada; Singapore	1,600,000	< fly	–	–
IMPALA	2018-02-05	DeepMind	United Kingdom	1,600,000	< fly	Confident	–
StyleGAN3-R	2021-06-21	Nvidia,Aalto University	Finland; United States	1,580,000	< fly	Confident	–
Deep Multitask NLP Network	2008-07-05	NEC Laboratories	United States	1,500,000	< fly	Speculative	unconfirmed
Double DQN	2016-03-02	Google DeepMind	United States	1,500,000	< fly	Confident	–
CNN committee (traffic sign)	2011-10-03	IDSIA	Switzerland	1,388,800	< fly	Likely	–
SqueezeNet	2016-02-24	DeepScale,University of California (UC) Berkeley,S	United States	1,200,000	< fly	–	–
Neocognitron	1980-04-01	NHK Broadcasting Science Research Laboratories	Japan	1,140,576	< fly	Confident	–
CURL	2020-04-08	University of California (UC) Berkeley	United States	907,264	< fly	–	–
DQN	2013-12-19	DeepMind	United Kingdom	836,096	< fly	Confident	–
EnhanceNet	2016-12-23	Max Planck Institute for Intelligent Systems	Germany	814,464	< fly	Confident	–
Image generation	2013-12-20	University of Amsterdam	Netherlands	784,000	< fly	Confident	–
Hanabi 4 player	2019-02-01	DeepMind,University of Oxford,Carnegie Mellon Univ	United Kingdom; United States	764,000	< fly	Confident	–
TA-CNN	2014-11-29	Chinese University of Hong Kong (CUHK)	Hong Kong	706,048	< fly	Likely	–
Robot Parkour	2023-09-12	Shanghai Qi Zhi institute,Stanford University,Carn	China; United States	500,000	< fly	Confident	–
LISSOM	1995-11-27	University of Texas at Austin	United States	432,800	< fly	Likely	–
SPIDER2	2016-10-28	Griffith University,University of Iowa,Dezhou Univ	Australia; China; United States	409,536	< fly	Likely	–
Mixture of linear models	1994-12-02			384,000	< fly	Likely	–
NPD	2014-08-06	IEEE	Multinational	313,856	< fly	Confident	–
Pooling CNN (Caltech 101)	2010-09-15	University of Bonn	Germany	294,912	< fly	Confident	–
Pooling CNN (NORB)	2010-09-15	University of Bonn	Germany	268,664	< fly	Confident	–
Two Stage Feature Extraction (MNIST)	2009-09-01	New York University (NYU)	United States	258,800	< fly	Confident	–
Vector Space Model	2011-06-19	Stanford University	United States	255,000	< fly	Confident	–
HyperNEAT	2014-03-05	University of Texas at Austin	United States	239,712	< fly	–	–
DNN EM segmentation	2012-12-03	IDSIA,SUPSI	Switzerland	218,896	< fly	Confident	–
Predictive Coding NN	1994-12-02	Technical University of Munich	Germany	206,910	< fly	Confident	–
Table Tennis Agent	2024-08-07	Google DeepMind	United States	185,000	< fly	Likely	–
CNN Committee (NIST)	2011-09-18	IDSIA	Switzerland	128,420	< fly	Likely	–
CNN Committee (MNIST)	2011-09-18	IDSIA	Switzerland	120,620	< fly	Likely	–
RECONTRA-uncategorized	1999-06-02			112,000	< fly	Likely	–
BLSTM for handwriting (2)	2007-12-03	University of Bern,IDSIA,Technical University of M	Germany; Switzerland	100,881	< fly	–	–
SVM-CNN	2006-06-17	New York University (NYU)	United States	90,857	< fly	Confident	–
Invariant CNN	2004-06-27	New York University (NYU)	United States	90,575	< fly	Confident	–
SexNet compression	1990-10-01			72,940	< fly	Confident	–
NeuroChess	1994-12-02			72,251	< fly	Speculative	unconfirmed
RECONTRA-categorized	1999-06-02			66,780	< fly	Likely	–
LeNet-5	1998-11-01	AT&T	United States	60,000	< fly	Confident	–
Swift	2023-08-30	Intel Labs	United States	56,804	< fly	Likely	–
TCN (P-MNIST)	2018-02-15	Carnegie Mellon University (CMU),Intel Labs	United States	42,000	< fly	Confident	–
TRPO	2015-02-19	University of California (UC) Berkeley	United States	33,500	< fly	Confident	–
SOM-CNN	1997-01-31			32,015	< fly	Confident	–
NETtalk reimplementation	1990-06-01	Oregon State University	United States	27,480	< fly	Confident	–
TD-Gammon	1992-05-01	IBM	United States	25,000	< fly	Speculative	unconfirmed
Cognitron	1975-09-01	Biological Cybernetics	Japan	21,600	< fly	Confident	–
NetTalk (transcription)	1987-06-06	Princeton University	United States	18,629	< fly	Confident	–
NetTalk (dictionary)	1987-06-06	Princeton University	United States	18,629	< fly	Confident	–
CFSS	2015-06-07	SenseTime,Chinese University of Hong Kong (CUHK),S	China; Hong Kong	17,408	< fly	Likely	–
MUSIC perceptron	1996-06-03			13,607	< fly	Confident	–
AdaRNN	2014-06-01	Beihang University	China	13,040	< fly	Confident	–
Bidirectional RNN	1997-11-01	Advanced Telecommunications Research Institute	Japan	13,000	< fly	–	–
Decision tree (classification)	2001-12-08	Mitsubishi Electric Research Labs,Compaq CRL	United States	12,000	< fly	Likely	–
LSTM	1997-11-15	Technical University of Munich	Germany	10,504	< fly	Confident	–
MLN-ASR	1988-08-01	McGill University	Canada	10,000	< fly	Confident	–
Hopfield network	1982-04-01	California Institute of Technology	United States	9,900	< fly	–	–
Zip CNN	1989-12-01	AT&T,Bell Laboratories	United States	9,760	< fly	Confident	–
Hierarchical Cognitron	1984-04-01	NHK Broadcasting Science Research Laboratories	Japan	9,315	< fly	Speculative	unconfirmed
Weight Decay	1991-12-02			8,386	< fly	Confident	–
Deep Blue	1997-05-01	IBM	United States	8,000	< fly	Speculative	unconfirmed
System 11	1996-06-18	Carnegie Mellon University (CMU)	United States	6,452	< fly	Confident	–
ACF-WIDER	2014-07-15	Chinese Academy of Sciences	China	6,144	< fly	Confident	–
RankNet	2005-08-07	Microsoft Research,Microsoft	United States	5,711	< fly	Confident	–
ANN Eye Tracker	1993-11-29			5,620	< fly	Confident	–
JPMAX	1994-12-02			4,446	< fly	Speculative	unconfirmed
Kohonen network	1981-07-25	Helsinki University of Technology	Finland	4,096	< fly	–	–
Speaker-independent vowel classification	1989-11-27	University of Washington	United States	3,040	< fly	Confident	–
Pattern recognition and reading by machine	1959-12-01	Sandia Corporation	United States	2,625	< fly	–	–
Handwritten digit recognition network	1989-11-27	AT&T	United States	2,578	< fly	Confident	–
Boosting	1992-11-30	Bell Laboratories	United States	2,578	< fly	Likely	–
Decision tree adaline	1969-05-01	Tokyo Medical and Dental University	Japan	2,450	< fly	Confident	–
LTE speaker verification system	1966-11-01	IBM	United States	2,061	< fly	Likely	–
Ceramic-MLP	1994-01-07	Sapienza Università di Roma	Italy	1,888	< fly	Likely	–
SexNet classification	1990-10-01			1,640	< fly	Likely	–
RAAM	1990-11-01			1,536	< fly	Confident	–
Super-vector coding	2010-01-01	University of Illinois Urbana-Champaign (UIUC),NEC	United States	1,025	< fly	Speculative	unconfirmed
Perceptron Mark I	1957-01-01	Cornell Aeronautical Laboratory,Cornell University	United States	1,000	< fly	Likely	–
Perceptron (1960)	1960-03-30	Cornell Aeronautical Laboratory	United States	1,000	< fly	Speculative	unconfirmed
Translation-invariant MLP	1987-06-15	Carnegie Mellon University (CMU)	United States	816	< fly	Confident	sparse (MoE)
Truck backer-upper	1989-06-18	Stanford University	United States	805	< fly	Confident	–
Siamese-TDNN	1993-08-01	Bell Laboratories	United States	744	< fly	Likely	–
MLP with back-propagation	1986-10-01	University of California San Diego,Carnegie Mellon	United States	720	< fly	Confident	–
Cancer drug mechanism prediction	1992-10-16	National Cancer Institute	United States	594	< fly	Likely	–
Distributed representation NN	1986-08-15	Carnegie Mellon University (CMU)	United States	432	< fly	Confident	sparse (MoE)
Piecewise linear model	1973-11-01	University of Kansas	United States	357	< fly	Confident	–
ASE+ACE	1983-09-01	University of Massachusetts Amherst	United States	324	< fly	Likely	–
LSTM with forget gates	1999-01-02	IDSIA	Switzerland	276	< fly	–	–
Self Organizing System	1955-03-01	Massachusetts Institute of Technology (MIT)	United States	225	< fly	–	–
Theseus	1950-07-02	Bell Laboratories	United States	40	< fly	Confident	–
SNARC	1952-01-08	Harvard University	United States	40	< fly	–	–
Bankruptcy-NN	1990-06-17			36	< fly	Confident	–
GNN	2008-12-09	University of Siena	Italy	30	< fly	Confident	–
ADALINE	1960-06-30	Stanford University	United States	17	< fly	Confident	–
Samuel Neural Checkers	1959-07-01	IBM	United States	16	< fly	Likely	–

16 How we built the dataset

1. 01

   Download

   Epoch AI’s notable-models CSV, 1,092 rows and 47 columns, downloaded on 10 November 2026 (CC BY 4.0); we saved the file and its checksum and read the licence from Epoch’s own pages ^{[1] [2]}.

2. 02

   Type the animals

   We typed each synapse count from a paper’s abstract or full text with the sentence it comes from, and a script found every sentence in the saved text (2371 checks in all). Neuron counts for the mammals come from our smartest-animals dataset and were re-read in the papers.

3. 03

   Calculate

   Where only neurons and synapses per neuron exist (cat, macaque, honeybee) the synapse count is neurons times synapses per neuron, shown as a range and labelled an IQ Metrics calculation. Nothing is filled in for a model with no count.

4. 04

   Compare

   For each animal we found the first row at or above its low and high count, the first real neural network, and the first year with three or more such models; we then counted models per tier and fitted the yearly maximum.

5. 05

   Publish

   The CSV holds all 1,092 model rows (size empty where Epoch has none), the 11 animals and 2 samples; every figure has its own CSV button.