Skip to content
Weights Not Walls
Menu

FrontierMath-Tier-4-v2-Private

Re-run by Epoch AI

The hardest tier of FrontierMath (second version): unpublished research-level mathematics problems written for Epoch AI.

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
83 in all: 16 for open-weight models, 67 for closed ones. 15 linked to a model page here.

Best open-weight result: moonshotai/Kimi-K3.1-Pro, 65.9%, number 27 of 83 results.

FrontierMath-Tier-4-v2-Private results
#ModelWeightsSettingScoreDateSource
1claude-fable-5-2_maxAnthropicclosedeffort max100.0%6 Nov 2026Epoch AI
2gpt-6.1-sol_maxOpenAIclosedeffort max100.0%29 Sep 2026Epoch AI
3gpt-6.2-astra_maxOpenAIclosedeffort max100.0%27 Oct 2026Epoch AI
4gpt-6-astra_highOpenAIclosedeffort high97.6%±2.430 Aug 2026Epoch AI
5gpt-6-astra_maxOpenAIclosedeffort max97.6%±2.430 Aug 2026Epoch AI
6gpt-6-astra_xhighOpenAIclosedeffort xhigh97.6%±2.430 Aug 2026Epoch AI
7gpt-6-astra_mediumOpenAIclosedeffort medium97.6%±2.430 Aug 2026Epoch AI
8gemini-4-carbon_highGoogle DeepMindclosedeffort high95.1%±3.42 Nov 2026Epoch AI
9claude-opus-5-5_maxAnthropicclosedeffort max95.0%±4.122 Sep 2026Epoch AI
10gpt-6.1-terra_maxOpenAIclosedeffort max92.7%±4.126 Oct 2026Epoch AI
11claude-fable-5_maxAnthropicclosedeffort max90.2%±4.69 Jun 2026Epoch AI log
12gpt-6-sol_maxOpenAIclosedeffort max90.0%±5.222 Sep 2026Epoch AI
13claude-fable-5-1_maxAnthropicclosedeffort max87.8%±5.21 Sep 2026Epoch AI
14gemini-4-hydrogen_xhighGoogle DeepMindclosedeffort xhigh87.8%±5.213 Nov 2026Epoch AI
15gpt-6-astra_lowOpenAIclosedeffort low87.8%±5.230 Aug 2026Epoch AI
16gpt-5.6-sol_maxOpenAIclosedeffort max82.9%±5.99 Jul 2026Epoch AI log
17gpt-6-astra_noneOpenAIclosedeffort none82.9%±5.930 Aug 2026Epoch AI
18claude-sonnet-5-5_maxAnthropicclosedeffort max80.5%±6.329 Sep 2026Epoch AI
19gpt-5.6-sol_promaxOpenAIcloseddefault80.5%±6.39 Jul 2026Epoch AI log
20gemini-4-argon_highGoogle DeepMindclosedeffort high78.0%±6.513 Oct 2026Epoch AI
21gpt-5.5-pro_xhighOpenAIclosedeffort xhigh78.0%±6.512 Jun 2026Epoch AI
22gdm-ai-co-mathematicianGoogle DeepMindcloseddefault75.6%±6.712 Jun 2026Epoch AI
23claude-opus-5_maxAnthropicclosedeffort max73.2%±7.024 Jul 2026Epoch AI log
24gpt-5.5_xhighOpenAIclosedeffort xhigh72.5%±7.111 Jun 2026Epoch AI
25gpt-5.6-terra_maxOpenAIclosedeffort max70.7%±7.29 Jul 2026Epoch AI log
26muse-wonder-1.0_maxMeta AIclosedeffort max70.7%±7.212 Oct 2026Epoch AI
27moonshotai/Kimi-K3.1-ProEpoch: kimi-k3.1-pro-together_max, Moonshotopeneffort max65.9%±7.56 Nov 2026Epoch AI
28gpt-5.6-luna_maxOpenAIclosedeffort max61.0%±7.79 Jul 2026Epoch AI log
29gpt-5.4-pro-2026-03-05_xhighOpenAIclosedeffort xhigh58.5%±7.813 Jun 2026Epoch AI log
30claude-opus-4-8_maxAnthropicclosedeffort max56.1%±7.810 Jun 2026Epoch AI log
31gpt-6-luna_maxOpenAIclosedeffort max56.1%±7.822 Sep 2026Epoch AI
32gpt-5.4-2026-03-05_xhighOpenAIclosedeffort xhigh49.0%±8.011 Jun 2026Epoch AI log
33claude-haiku-5-5_maxAnthropicclosedeffort max46.3%±7.97 Oct 2026Epoch AI
34deepseek-ai/DeepSeek-V4.1-Flash-1017Epoch: deepseek-v4.1-flash-1017_max, DeepSeekopeneffort max46.3%±7.919 Oct 2026Epoch AI
35muse-spark-1.3_maxMeta AIclosedeffort max46.3%±7.918 Sep 2026Epoch AI
36qwen3.8-max_xhighAlibabaclosedeffort xhigh46.3%±7.94 Aug 2026Epoch AI log
37gpt-5.2-pro-2025-12-11_xhighOpenAIclosedeffort xhigh46.0%±7.813 Jun 2026Epoch AI
38muse-spark-1.3_xhighMeta AIclosedeffort xhigh41.5%±7.816 Sep 2026Epoch AI
39moonshotai/Kimi-K3Epoch: kimi-k3_max, Moonshotopeneffort max39.0%±7.717 Jul 2026Epoch AI log
40gemini-3.7-flash_highGoogle DeepMindclosedeffort high36.6%±7.614 Aug 2026Epoch AI log
41qwen3.7-maxAlibabacloseddefault34.1%±7.513 Jun 2026Epoch AI log
42qwen3.8-max-0902_xhighAlibabaclosedeffort xhigh34.1%±7.52 Sep 2026Epoch AI
43claude-opus-4-7_maxAnthropicclosedeffort max31.7%±7.410 Jun 2026Epoch AI log
44grok-4.6_xhighxAIclosedeffort xhigh31.7%±7.414 Aug 2026Epoch AI log
45meta-models/Muse-Spark-Open-816B-A42BEpoch: Muse-Spark-Open-816B-A42B_xhigh, Meta AIopeneffort xhigh31.7%±7.415 Oct 2026Epoch AI
46gpt-5.2-2025-12-11_xhighOpenAIclosedeffort xhigh31.7%±7.311 Jun 2026Epoch AI
47claude-sonnet-5_maxAnthropicclosedeffort max29.3%±7.230 Jun 2026Epoch AI log
48zai-org/GLM-5.2Epoch: glm-5.2_max, Z.ai (Zhipu AI)openeffort max29.3%±7.219 Jun 2026Epoch AI log
49zai-org/GLM-5.3Epoch: glm-5.3_max, Z.ai (Zhipu AI)openeffort max29.3%±7.225 Aug 2026Epoch AI
50claude-opus-4-6_maxAnthropicclosedeffort max26.8%±7.011 Jun 2026Epoch AI log
51deepseek-ai/DeepSeek-V4-Pro-0813Epoch: deepseek-v4-pro-0813_max, DeepSeekopeneffort max26.8%±7.019 Aug 2026Epoch AI log
52deepseek-ai/DeepSeek-V4.1-FlashEpoch: deepseek-v4.1-flash_max, DeepSeekopeneffort max26.8%±7.08 Oct 2026Epoch AI
53gemini-3.1-pro-previewGoogle DeepMindcloseddefault26.8%±7.011 Jun 2026Epoch AI log
54gemini-3.5-flash_highGoogle DeepMindclosedeffort high26.8%±7.010 Jun 2026Epoch AI log
55mistral-large-4-0_highMistral AIno page matchopeneffort high26.8%±7.012 Oct 2026Epoch AI
56moonshotai/Kimi-K2.6Epoch: kimi-k2.6, Moonshotopendefault25.6%±7.110 Jun 2026Epoch AI log
57deepseek-ai/DeepSeek-V4-Flash-0731Epoch: deepseek-v4-flash-0731_max, DeepSeekopeneffort max24.4%±6.82 Aug 2026Epoch AI log
58grok-4.5_highxAIclosedeffort high24.4%±6.89 Jul 2026Epoch AI
59gemini-3.6-flash_highGoogle DeepMindclosedeffort high22.0%±6.52 Aug 2026Epoch AI log
60gemini-3.8-flash_highGoogle DeepMindclosedeffort high22.0%±6.52 Sep 2026Epoch AI
61gpt-5-2025-08-07_highOpenAIclosedeffort high22.0%±6.511 Jun 2026Epoch AI
62gpt-5-pro-2025-10-06_highOpenAIclosedeffort high19.5%±6.312 Jun 2026Epoch AI log
63gemini-3-flash-previewGoogle DeepMindcloseddefault17.1%±5.911 Jun 2026Epoch AI log
64zai-org/GLM-5.3-FlashEpoch: glm-5.3-flash_max, Z.ai (Zhipu AI)openeffort max17.1%±5.927 Aug 2026Epoch AI
65grok-4.20-0309-reasoningxAIcloseddefault17.1%±5.913 Jul 2026Epoch AI log
66grok-4.7_xhighxAIclosedeffort xhigh17.1%±5.922 Sep 2026Epoch AI
67thinkingmachines/Inkling-SmallEpoch: inkling-small_xhigh, Thinking Machinesopeneffort xhigh17.1%±5.914 Aug 2026Epoch AI log
68grok-4.3_highxAIclosedeffort high14.6%±5.617 Jun 2026Epoch AI log
69gpt-5-mini-2025-08-07_highOpenAIclosedeffort high12.2%±5.212 Jun 2026Epoch AI log
70gpt-5.4-nano-2026-03-17_highOpenAIclosedeffort high12.2%±5.212 Jun 2026Epoch AI log
71moonshotai/Kimi-K2.7-CodeEpoch: kimi-k2.7-code, Moonshotopendefault12.2%±5.213 Jun 2026Epoch AI log
72gpt-5.4-mini-2026-03-17_xhighOpenAIclosedeffort xhigh9.8%±4.712 Jun 2026Epoch AI
73claude-opus-4-5-20251101_32KAnthropiccloseddefault4.9%±3.411 Jun 2026Epoch AI log
74thinkingmachines/InklingEpoch: Inkling_xhigh, Thinking Machinesopeneffort xhigh4.9%±3.46 Aug 2026Epoch AI log
75o4-mini-2025-04-16_highOpenAIclosedeffort high4.9%±3.411 Jun 2026Epoch AI log
76claude-opus-4-1-20250805_32KAnthropiccloseddefault2.4%±2.411 Jun 2026Epoch AI log
77claude-sonnet-4-5-20250929_32KAnthropiccloseddefault2.4%±2.411 Jun 2026Epoch AI log
78deepseek-ai/DeepSeek-V4-ProEpoch: deepseek-v4-pro_max, DeepSeekopeneffort max2.4%±2.417 Jun 2026Epoch AI log
79gpt-5-nano-2025-08-07_highOpenAIclosedeffort high2.4%±2.412 Jun 2026Epoch AI log
80gpt-5.5-instantOpenAIcloseddefault2.4%±2.42 Aug 2026Epoch AI log
81gemini-2.5-proGoogle DeepMindcloseddefault0.0%11 Jun 2026Epoch AI log
82gemini-3.5-flash-lite_highGoogle DeepMindclosedeffort high0.0%2 Aug 2026Epoch AI log
83o3-mini-2025-01-31_highOpenAIclosedeffort high0.0%11 Jun 2026Epoch AI log

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