Skip to content
Weights Not Walls
Menu

EBR-bench

Re-run by Epoch AI

No description yet: we only describe an eval set when we can do it correctly.

Kind
Run independently by Epoch AI, with public logs. Our own re-runs are not built yet.
Score shown
Epoch's column "Best score (across scorers)", as a percentage, higher is better; random guessing scores about 0.0%. Error bars as the source gives them.
Results
33 in all: 3 for open-weight models, 30 for closed ones. 2 linked to a model page here.

Best open-weight result: mistral-large-4-0-oss_high, 18.1%, number 23 of 33 results.

EBR-bench results
#ModelWeightsSettingScoreDateSource
1gpt-6.2-astra_maxOpenAIclosedeffort max92.4%±4.927 Oct 2026Epoch AI
2claude-fable-5-2_maxAnthropicclosedeffort max90.5%±1.56 Nov 2026Epoch AI
3gemini-4-carbon_highGoogle DeepMindclosedeffort high87.6%±6.85 Nov 2026Epoch AI
4gpt-6.1-terra_maxOpenAIclosedeffort max78.1%±5.126 Oct 2026Epoch AI
5gpt-6-astra_maxOpenAIclosedeffort max76.2%±8.430 Aug 2026Epoch AI
6claude-opus-5-5_maxAnthropicclosedeffort max71.4%±7.122 Sep 2026Epoch AI
7gemini-4-hydrogen_xhighGoogle DeepMindclosedeffort xhigh70.5%±6.313 Nov 2026Epoch AI
8gemini-4-argon_highGoogle DeepMindclosedeffort high64.8%±4.913 Oct 2026Epoch AI
9claude-fable-5-1_maxAnthropicclosedeffort max57.1%±6.08 Sep 2026Epoch AI
10gpt-6.1-sol_maxOpenAIclosedeffort max54.3%±5.829 Sep 2026Epoch AI
11gpt-6-sol_maxOpenAIclosedeffort max53.3%±4.622 Sep 2026Epoch AI
12claude-opus-5_maxAnthropicclosedeffort max45.7%±3.65 Sep 2026Epoch AI
13gpt-5.6-sol_maxOpenAIclosedeffort max44.8%±4.24 Sep 2026Epoch AI
14claude-haiku-5-5_maxAnthropicclosedeffort max43.8%±1.89 Oct 2026Epoch AI
15claude-fable-5_maxAnthropicclosedeffort max39.5%±2.017 Jul 2026Epoch AI
16gpt-5.5_xhighOpenAIclosedeffort xhigh34.3%±2.027 Jul 2026Epoch AI
17muse-wonder-1.0_maxMeta AIclosedeffort max31.4%±4.212 Oct 2026Epoch AI
18grok-4.6_xhighxAIclosedeffort xhigh30.5%±1.618 Aug 2026Epoch AI
19claude-opus-4-8_maxAnthropicclosedeffort max28.6%±1.47 Aug 2026Epoch AI
20gpt-5.4-2026-03-05_xhighOpenAIclosedeffort xhigh25.4%25 Jun 2026Epoch AI
21gpt-5.2-2025-12-11_xhighOpenAIclosedeffort xhigh23.0%26 Jun 2026Epoch AI
22claude-opus-4-7_maxAnthropicclosedeffort max19.0%30 Jun 2026Epoch AI
23mistral-large-4-0-oss_highMistral AIno page matchopeneffort high18.1%±3.22 Nov 2026Epoch AI
24claude-opus-4-5-20251101_128KAnthropiccloseddefault14.3%25 Jun 2026Epoch AI
25gemini-3.1-pro-previewGoogle DeepMindcloseddefault14.3%25 Jun 2026Epoch AI
26claude-opus-4-6_maxAnthropicclosedeffort max12.7%29 Jun 2026Epoch AI
27gpt-5-2025-08-07_highOpenAIclosedeffort high12.7%29 Jun 2026Epoch AI
28zai-org/GLM-5.2Epoch: glm-5.2_max, Z.ai (Zhipu AI)openeffort max9.5%29 Jun 2026Epoch AI
29qwen3.7-maxAlibabacloseddefault9.5%25 Jun 2026Epoch AI
30claude-opus-4-1-20250805Anthropiccloseddefault7.9%25 Jun 2026Epoch AI
31gemini-3.5-flash_highGoogle DeepMindclosedeffort high4.8%25 Jun 2026Epoch AI
32claude-sonnet-4-5-20250929Anthropiccloseddefault2.4%25 Jun 2026Epoch AI
33moonshotai/Kimi-K2.6Epoch: kimi-k2.6, Moonshotopendefault2.4%25 Jun 2026Epoch AI

Open-weight rows are shaded. Dates are when the result was run or evaluated, where the data says. Data from Epoch AI (CC-BY 4.0), fetched 19 Nov 2026: Epoch AI, ‘Capabilities & benchmarking’. Published online at epoch.ai. Retrieved from ‘https://epoch.ai/benchmarks’ [online resource]. Epoch AI. Method