Get started

Epoch Capabilities Index

One number per model, fit by Epoch over the benchmarks that model has results on, with the interval around it. 258 models placed, 116 of them listed in the catalog Vitna prices.

retrieved ·CC-BY-4.0epoch.ai/eci ↗

Models placed258each on 4 benchmark results or more
Results behind it2,917counted here, not republished
Scale130 / 150anchored at Claude 3.5 Sonnet and GPT-5
Released2023 to 2026by the date each model came out

What the interval says

TOP 24 OF 258

Claude Fable 5.2 leads at 173.4, and 8 other models have intervals reaching its own, so the index does not separate them.

  1. 01Claude Fable 5.2Anthropic173.4
  2. 02GPT-6.2 AstraOpenAI172.8
  3. 03Gemini 4 CarbonGoogle DeepMind169.7
  4. 04GPT-6 AstraOpenAI169.2
  5. 05Claude Opus 5.5Anthropic168.7
  6. 06GPT-6.1 SolOpenAI168.6
  7. 07GPT-6.1 TerraOpenAI167.7
  8. 08Claude Fable 5.1Anthropic167.1
  9. 09Gemini 4 HydrogenGoogle DeepMind166.8
  10. 10Claude Sonnet 5.5Anthropic166.6
  11. 11Gemini 4 ArgonGoogle DeepMind166.5
  12. 12Claude Opus 5Anthropic164.7
  13. 13Claude Fable 5Anthropic164.2
  14. 14GPT-6 SolOpenAI163.8
  15. 15GPT-5.6 SolOpenAI163.0
  16. 16GPT-5.5 ProOpenAI162.8
  17. 17Muse WonderMeta AI161.7
  18. 18Kimi K3.1 ProMoonshot161.2
  19. 19GPT-5.6 TerraOpenAI160.8
  20. 20Claude Haiku 5.5Anthropic160.3
  21. 21GPT-5.4 ProOpenAI160.3
  22. 22GPT-5.5OpenAI160.2
  23. 23Claude Opus 4.8Anthropic159.9
  24. 24DeepSeek V4.1 Flash 1017DeepSeek159.4

The rail runs 157.1 to 177.2, the range these 24 rows cover rather than the whole index, because scaled to all 258 they would sit on top of one another. Each bar is the interval Epoch publishes and the mark is the figure; a marked row is one whose interval reaches that of the leader. Epoch states that absolute values are meaningless by themselves and that comparisons between models are what the index is for, that values are not linearly related to benchmark accuracy, and that a highly specialised model may score low while being capable in its own domain.

Vitna fits a scale of its own over the six openly licensed boards, 319 ranked runs over 46 columns, and this is the published index the ordering of that fit is checked against. The two are separate fits over different tables, so their numbers do not compare and only their orderings do.

113 of 258.

Index figures, as Epoch publishes them
#ModelOrganisationWeightsReleasedResultsIntervalIndex
18Kimi K3.1 ProMoonshotOpen weights16158.9 to 163.7161.2
24DeepSeek V4.1 Flash 1017DeepSeekOpen weights11157.1 to 161.9159.4
26Kimi K3MoonshotOpen weights22156.7 to 160.8158.6
34GLM-5.3Z.ai (Zhipu AI)Open weights16155.0 to 159.4157.1
35DeepSeek V4.1 FlashDeepSeekOpen weights20155.1 to 159.2157.0
43DeepSeek V4 Pro 0813DeepSeekOpen weights16154.1 to 158.1156.0
44Muse Spark OpenMeta AIOpen weights8153.3 to 157.8155.5
50Mistral Large 4Mistral AIOpen weights7151.8 to 157.4154.6
53DeepSeek V4 Flash 0731DeepSeekOpen weights15152.6 to 156.4154.4
54GLM-5.3-FlashZ.ai (Zhipu AI)Open weights14151.9 to 155.8153.8
56GLM-5.2Z.ai (Zhipu AI)Open weights20151.3 to 154.9153.1
59Mistral Large 4 PreviewMistral AIOpen weights16150.8 to 154.7152.7
60Qwen 3.8 27BAlibabaOpen weights9150.7 to 154.7152.6
64Kimi K2.6MoonshotOpen weights15149.9 to 153.4151.6
66Kimi K2.7 CodeMoonshotOpen weights11149.2 to 153.3151.2
68Beam 1.0ReflectionOpen weights4148.6 to 153.1150.8
69Inkling-SmallThinking MachinesOpen weights9149.0 to 152.5150.8
70GLM-5.1Z.ai (Zhipu AI)Open weights11148.7 to 152.8150.7
71DeepSeek-V4-ProDeepSeekOpen weights14148.4 to 152.0150.2
77Kimi K2.5MoonshotOpen weights14147.3 to 150.3148.8
78InklingThinking MachinesOpen weights17146.8 to 150.1148.4
79Qwen3.5 397B-A17BAlibabaOpen weights7145.9 to 150.1148.0
81MiniMax-M3MiniMaxOpen weights12145.7 to 150.0147.8
85Nemotron 3 UltraNvidiaOpen weights10144.9 to 149.2147.0
87DeepSeek-V4-FlashDeepSeekOpen weights4143.7 to 150.2146.9
89Qwen3.6 27BAlibabaOpen weights6144.6 to 148.9146.7
90GLM-5Z.ai (Zhipu AI)Open weights10144.9 to 148.1146.5
92DeepSeek-V3.2DeepSeekOpen weights14144.7 to 147.8146.3
95Kimi K2 ThinkingMoonshotOpen weights9143.7 to 147.3145.5
100DeepSeek-V3.2-ExpDeepSeekOpen weights8142.2 to 146.7144.4
102Gemma 4 31B ITGoogle DeepMindOpen weights7141.7 to 146.6144.1
103Qwen 3.6 35B-A3BAlibabaOpen weights9142.2 to 145.9144.1
104Qwen3.5-35B-A3BAlibabaOpen weights4141.5 to 146.6144.1
108GLM-4.7Z.ai (Zhipu AI)Open weights11141.4 to 145.3143.3
109Qwen3-235B-A22B-Thinking (Jul 2025)AlibabaOpen weights10141.3 to 145.1143.2
115Mistral Medium 3.5Mistral AIOpen weights6138.3 to 144.7141.6
117Gemma 4 26B A4BGoogle DeepMindOpen weights5139.1 to 143.9141.6
118Kimi K2 (Jul 2025)MoonshotOpen weights6138.4 to 143.9141.3
121DeepSeek-V3.1DeepSeekOpen weights5137.7 to 144.0140.9
123DeepSeek-R1 (May 2025)DeepSeekOpen weights11138.9 to 142.4140.7
127gpt-oss-120bOpenAIOpen weights13138.3 to 142.0140.2
129Qwen3.5-9BAlibabaOpen weights5137.3 to 141.7139.5
131DeepSeek-R1DeepSeekOpen weights11136.6 to 140.5138.6
132Qwen3-235B-A22BAlibabaOpen weights8136.0 to 140.7138.4
133Qwen3-30B-A3B-Thinking (Jul 2025)AlibabaOpen weights4135.8 to 140.8138.3
134Qwen3-235B-A22B-Instruct (Jul 2025)AlibabaOpen weights7135.7 to 140.2138.0
136Qwen3-32BAlibabaOpen weights6134.4 to 139.1136.8
138QwQ-32BAlibabaOpen weights4133.7 to 139.1136.4
140Qwen3-14BAlibabaOpen weights5133.3 to 138.7136.1
143DeepSeek-V3 (Mar 2025)DeepSeekOpen weights10133.0 to 137.4135.3
144gpt-oss-20bOpenAIOpen weights6132.3 to 137.3134.9
145Qwen3-30B-A3B-Instruct (Jul 2025)AlibabaOpen weights4131.9 to 137.7134.9
147Qwen3-30B-A3BAlibabaOpen weights6131.3 to 136.7134.2
154Llama 4 MaverickMeta AIOpen weights13129.8 to 134.3132.2
155Qwen3-8BAlibabaOpen weights5128.9 to 134.9132.1
156DeepSeek-V3DeepSeekOpen weights14129.2 to 134.0131.7
161Phi-4Microsoft ResearchOpen weights5126.1 to 132.0129.2
164Llama 3.1-405BMeta AIOpen weights15125.0 to 130.5127.9
165Llama 4 ScoutMeta AIOpen weights9124.9 to 130.3127.8
166Gemma 3 27BGoogle DeepMindOpen weights8124.4 to 130.4127.6
167Qwen2.5-72BAlibabaOpen weights18124.4 to 129.8127.2
171Mistral Large 2 (Jul 2024)Mistral AIOpen weights6123.1 to 129.4126.4
172Llama 3.3 70BMeta AIOpen weights10123.0 to 129.0126.2
178Llama 3.1-70BMeta AIOpen weights9121.4 to 127.7124.7
180Llama 3.2 90BMeta AIOpen weights6119.2 to 126.3122.9
181Qwen2-72BAlibabaOpen weights6118.6 to 125.6122.3
183Llama 3-70BMeta AIOpen weights7117.5 to 124.7121.3
184DeepSeek-V2 (MoE-236B, May 2024)DeepSeekOpen weights7114.0 to 127.6120.8
189Qwen2.5-Coder-32BAlibabaOpen weights5109.5 to 127.8118.7
190Mixtral 8x22BMistral AIOpen weights6114.5 to 122.5118.7
191Gemma 2 27BGoogle DeepMindOpen weights6114.2 to 122.4118.6
194phi-3-small 7.4BMicrosoftOpen weights8110.5 to 121.9116.3
195phi-3-medium 14BMicrosoftOpen weights10109.2 to 118.5114.1
198Mistral NeMoMistral AIOpen weights6108.6 to 118.4113.9
200Mixtral 8x7BMistral AIOpen weights12108.1 to 117.5113.2
203Gemma 2 9BGoogle DeepMindOpen weights6106.8 to 117.3112.4
204Llama 3.1-8BMeta AIOpen weights10105.1 to 115.9110.9
205phi-3-mini 3.8BMicrosoftOpen weights8104.0 to 116.1110.5
206Stable Beluga 2Stability AIOpen weights7102.0 to 116.1109.4
207Llama 3-8BMeta AIOpen weights10103.4 to 114.0109.1
208Qwen2.5-Coder (7B)AlibabaOpen weights597.7 to 116.1107.4
209Falcon-180BTechnology Innovation InstituteOpen weights898.2 to 115.5107.0
211Llama 2-70BMeta AIOpen weights1499.7 to 111.0105.7
212Gemma 7BGoogle DeepMindOpen weights1095.4 to 109.7103.2
213Mistral 7B v0.1Mistral AIOpen weights1094.9 to 109.1102.7
215Qwen-14BAlibabaOpen weights693.2 to 109.3101.8
216LLaMA-65BMeta AIOpen weights1093.2 to 108.9101.5
218Falcon 2 11BTechnology Innovation InstituteOpen weights488.4 to 109.099.2
223LLaMA-33BMeta AIOpen weights1087.4 to 104.596.5
224Llama 2-13BMeta AIOpen weights1187.4 to 104.296.5
225Phi-2MicrosoftOpen weights886.9 to 104.396.2
226StarCoder 2 15BHugging Face,ServiceNow,Nvidia,BigCodeOpen weights484.0 to 106.096.1
227Yi 6B01.AIOpen weights684.1 to 103.294.3
230Baichuan2-13BBaichuanOpen weights780.8 to 100.991.8
231Falcon-40BTechnology Innovation InstituteOpen weights1080.0 to 100.090.7
232Qwen-7BAlibabaOpen weights677.9 to 98.789.1
233Qwen2.5-Coder (1.5B)AlibabaOpen weights576.0 to 100.189.1
234LLaMA-13BMeta AIOpen weights1073.4 to 95.485.2
235MPT-30BMosaicMLOpen weights873.3 to 95.385.2
237Baichuan 2-7BBaichuanOpen weights670.5 to 94.583.6
239Llama 2-7BMeta AIOpen weights1070.6 to 93.883.2
240LLaMA-7BMeta AIOpen weights1063.4 to 88.977.3
241DeepSeek Coder 33BDeepSeek,Peking UniversityOpen weights460.4 to 89.576.3
243Gemma 2BGoogle DeepMindOpen weights861.0 to 87.675.7
246MPT-7BMosaicMLOpen weights1058.2 to 86.273.4
247Falcon-7BTechnology Innovation InstituteOpen weights1058.2 to 85.473.1
248StarCoder 2 7BHugging Face,ServiceNow,Nvidia,BigCodeOpen weights454.9 to 86.672.4
249XGen-7BSalesforceOpen weights654.3 to 84.270.3
250Phi-1.5MicrosoftOpen weights550.5 to 82.968.4
252DeepSeek Coder 6.7BDeepSeek,Peking UniversityOpen weights449.1 to 82.267.5
253StarCoder 2 3BHugging Face,ServiceNow,Nvidia,BigCodeOpen weights447.9 to 81.766.6
257Dolly 2.0-12bDatabricksOpen weights634.5 to 72.655.5
258Cerebras-GPT-13BCerebras SystemsOpen weights626.3 to 67.148.9

113 of 258 models, highest index first. A model links to its catalog page where the pairing names one, which is true of 116; the rest are models the catalog does not list, and the record Epoch keeps of them stands on its own. Results is how many benchmark results stand behind the figure, and Epoch places nothing on fewer than 4. The two rows with no interval are the anchors the scale is defined by, whose figures are fixed rather than estimated.

Sources

Source: the Epoch AI Capabilities Index (epoch.ai/eci), as of 20 Nov 2026. 258 models placed over 2917 benchmark results.
Epoch AI Benchmarking Hub data, by Epoch AI, under CC BY 4.0.
Epoch states that absolute values are meaningless by themselves and that comparisons between models are what the index is for, that values are not linearly related to benchmark accuracy, and that a highly specialised model may score low while being capable in its own domain.
The index is Epoch's fit. Vitna ran none of these evaluations, computed none of these figures and adjusted none of them.