FrontierMath v2 Tiers 1–3

Epoch AI · FrontierMath v2 Tiers 1–3 的公开模型评测结果。

官方评测

来源说明

该来源测量模型在「推理」领域的具体交付表现，原始成绩会和统一能力读数分开展示。

限制: 不同版本、推理设置和测试集可能导致分数变化，请结合数据时间阅读。

使用方式

参与排名的来源会影响能力分；参考、交叉参考和观察来源仅用于对比与核验。

本来源的模型成绩


GPT-6 Astra	0.93684211	72.8	2026-11-05
GPT-6.1 Sol	0.93684211	72.8	2026-11-05
Claude Opus 5.5	0.9122807	71.2	2026-11-05
GPT-6.1 Terra	0.90877193	71.0	2026-11-05
Claude Fable 5.1	0.90175439	70.6	2026-11-05
GPT-6 Sol	0.89824561	70.4	2026-11-05
GPT-5.6 Sol	0.89122807	69.9	2026-11-05
Claude Sonnet 5.5	0.8877193	69.7	2026-11-05
GPT-5.5 Pro	0.87719298	69.1	2026-11-05
Claude Fable 5	0.87017544	68.6	2026-11-05
GPT-5.6 Terra	0.85964912	68.0	2026-11-05
Gemini 4 Argon	0.85964912	68.0	2026-10-17
Gemini 4 Argon	0.85964912	68.0	2026-11-05
Gemini 4 Argon	0.85964912	68.0	2026-10-18
Claude Opus 5	0.85614035	67.7	2026-11-05
GPT-5.5	0.85263158	67.5	2026-11-05
GPT-5.4 Pro	0.8245614	65.8	2026-11-05
GPT-5.6 Luna	0.82105263	65.6	2026-11-05
Muse Wonder	0.82105263	65.6	2026-11-05
Muse Wonder	0.82105263	65.6	2026-10-18
Muse Wonder	0.82105263	65.6	2026-10-17
Claude Opus 4.8	0.8	64.2	2026-11-05
GPT-6 Luna	0.78947368	63.6	2026-11-05
GPT-5.4	0.78596491	63.4	2026-11-05
Claude Haiku 5.5	0.75087719	61.2	2026-11-05
Qwen3.8 Max	0.74736842	61.0	2026-11-05
Muse Spark 1.3	0.74385965	60.8	2026-11-05
GPT-5.2 Pro	0.74	60.5	2026-11-05
Kimi K3	0.72183099	59.4	2026-11-05
DeepSeek V4.1 Flash 1017	0.7122807	58.8	2026-11-05
Claude Opus 4.7	0.70175439	58.1	2026-11-05
GLM 5.3	0.6877193	57.3	2026-11-05
Gemini 3.8 Flash	0.68421053	57.0	2026-11-05
GPT-5.2	0.674	56.4	2026-11-05
DeepSeek V4.1 Flash	0.67368421	56.4	2026-11-05
Claude Opus 4.6	0.65964912	55.5	2026-11-05
Grok 4.6	0.65964912	55.5	2026-11-05
Claude Sonnet 5	0.65614035	55.3	2026-11-05
Qwen3.7 Max	0.64561404	54.6	2026-11-05
DeepSeek V4 Pro 0813	0.64561404	54.6	2026-11-05
Gemini 3.1 Pro Preview	0.59649123	51.6	2026-11-05
GLM-5.2	0.59205776	51.3	2026-11-05
Gemini 3.6 Flash	0.58947368	51.2	2026-11-05
Mistral Large 4	0.58596491	50.9	2026-10-31
Mistral Large 4	0.58596491	50.9	2026-10-17
Mistral Large 4	0.58596491	50.9	2026-10-18
DeepSeek V4 Flash 0731	0.5754386	50.3	2026-11-05
Kimi K2.6	0.57192982	50.1	2026-11-05
Grok 4.5	0.57192982	50.1	2026-11-05
GPT-5 Pro	0.55789474	49.2	2026-11-05
GLM-5.3 Flash	0.55789474	49.2	2026-10-18
GLM-5.3 Flash	0.55789474	49.2	2026-11-05
GLM-5.3 Flash	0.55789474	49.2	2026-10-17
GPT-5	0.55438596	49.0	2026-11-05
Kimi K2.7 Code	0.54035088	48.1	2026-11-05
Grok 4.7	0.52982456	47.4	2026-11-05
Gemini 3 Flash Preview	0.5122807	46.4	2026-11-05
GPT-5.4 mini	0.5122807	46.4	2026-11-05
GPT-5 mini	0.46666667	43.5	2026-11-05
Inkling Small	0.46315789	43.3	2026-11-05
Grok 4.20	0.44912281	42.4	2026-11-05
Qwen3.6 Plus	0.38245614	38.3	2026-11-05
GLM-5.1	0.36842105	37.4	2026-11-05
o4 Mini High	0.36140351	37.0	2026-11-05
Qwen3.6 27B	0.35087719	36.3	2026-11-05
Claude Opus 4.5	0.34385965	35.9	2026-11-05
Qwen3.7 Plus	0.34385965	35.9	2026-11-05
o3	0.33333333	35.2	2026-11-05
Inkling	0.33333333	35.2	2026-11-05
Qwen3.5 397B-A17B	0.3122807	33.9	2026-11-05
o4-mini	0.2877193	32.4	2026-11-05
Gemini 3.1 Flash Lite	0.27719298	31.7	2026-11-05
GPT-5.5 Instant	0.26315789	30.9	2026-11-05
Gemini 3.5 Flash Lite	0.25964912	30.6	2026-11-05
Gemini 2.5 Pro	0.24561404	29.8	2026-11-05
Claude Sonnet 4.5	0.23859649	29.3	2026-11-05
Qwen3.6 Flash	0.2245614	28.5	2026-11-05
Qwen3.6 35B-A3B	0.20350877	27.2	2026-11-05
GPT-5 nano	0.2	26.9	2026-11-05
Qwen3.7 Flash	0.19298246	26.5	2026-11-05
Qwen3 Max	0.18947368	26.3	2026-11-05
o3 Mini High	0.18596491	26.1	2026-11-05
Qwen3.5-Flash	0.18245614	25.8	2026-11-05
o1	0.14736842	23.7	2026-11-03
o3-mini	0.10526316	21.0	2026-11-05

FrontierMath v2 Tiers 1–3

本来源的模型成绩

常见问题

如何核对 FrontierMath v2 Tiers 1–3 的成绩？

本页的原始成绩会直接成为综合能力分吗？

为什么这个来源没有覆盖所有模型？