Pass IndexThe State of AISign in

Live bench — Epoch AI

Live bench — Epoch AI is run by Epoch AI. It has ranked 54 models, of which the catalogue holds 37, scoring from 22.36 to 78.79.

measured by Epoch AI · the board itself

PlaceModelMetricScore
2ndof 54GPT-5.1Global average (high)78.79
3rdof 54Claude 3.7 SonnetGlobal average76.1
4thof 54o3-miniGlobal average (high)75.88
5thof 54o1Global average (high)75.67
6thof 54QwQ-32BGlobal average71.96
7thof 54DeepSeek-R1Global average71.57
11thof 54DeepSeek-V3 0324Global average66.86
16thof 54GPT-5Global average (max)62.29
17thof 54gemini-2.0-flash-001Global average61.47
18thof 54DeepSeek-V3Global average60.45
19thof 54Gemini 2.0 FlashGlobal average59.26
20thof 54Claude 3.5 SonnetGlobal average59.03
21stof 54o1-miniGlobal average (medium)57.76
22ndof 54GPT-4o (2024-08-06)Global average55.33
23rdof 54R1 Distill Llama 70BGlobal average54.5
25thof 54Gemini 2.0 Flash-LiteGlobal average54.29
28thof 54GPT-4o (2024-11-20)Global average52.19
30thof 54Llama 3.3 70B InstructGlobal average50.16
31stof 54Gemma 3 27BGlobal average49.99
32ndof 54Claude 3 OpusGlobal average49.16
34thof 54SonarGlobal average46.88
35thof 54Qwen2.5 Coder 32B InstructGlobal average46.23
37thof 54DeepSeek R1 Distill QWEN 32BGlobal average45.55
38thof 54mistral-small-2503Global average43.96
39thof 54Nova ProGlobal average43.53
40thof 54Claude Haiku 3.5Global average43.45
42ndof 54Phi-4Global average41.61
43rdof 54GPT-4o-mini (2024-07-18)Global average41.26
44thof 54QwQ-32B-PreviewGlobal average40.25
45thof 54Gemma 2 27BGlobal average38.18
46thof 54Nova LiteGlobal average36.35
47thof 54c4ai-command-r-plus-08-2024Global average31.76
48thof 54Nova MicroGlobal average29.59
49thof 54gemma-2-9b-itGlobal average28.66
50thof 54c4ai-command-r-08-2024Global average27.48
51stof 54Phi-3-small-8k-instructGlobal average24.03
52ndof 54Phi-3-mini-4k-instructGlobal average22.36

Read from the board on 2026-09-16