Benchmark

FrontierMath Tier 4

Measures: 43 exceptionally hard, research-level unpublished math problems (the FrontierMath expansion tier) with automatically verifiable answers; 1 point per correct answer on the private set.

Maintainer
Epoch AI (OpenAI-funded; conflict of interest disclosed)
Introduced
2025
Status
Active
Saturation
Saturated
Typical provenance
Independent
In scoring set
No

PaperOfficial leaderboard

Leaderboard

Independent and third-party results only.

Leaderboard
ModelAccuracyProvenanceSourceEvaluated
GPT-6 Astra
OpenAI · High effort
97.6%IndependentEpoch AI30 Aug 2026
GPT-6 Astra
OpenAI · Xhigh effort
97.6%IndependentEpoch AI30 Aug 2026
GPT-6 Astra
OpenAI · Max effort
97.6%IndependentEpoch AI30 Aug 2026
GPT-6 Astra
OpenAI · Medium effort
97.6%IndependentEpoch AI30 Aug 2026
Claude Opus 5.5
Anthropic · Max effort
95.0%IndependentEpoch AI22 Sep 2026
Claude Fable 5
Anthropic · Max effort
90.2%IndependentEpoch AI9 Jun 2026
Claude Fable 5.1
Anthropic · Max effort
87.8%IndependentEpoch AI1 Sep 2026
GPT-6 Astra
OpenAI · Low effort
87.8%IndependentEpoch AI30 Aug 2026
GPT-5.6 Sol
OpenAI · Max effort
82.9%IndependentEpoch AI9 Jul 2026
GPT-6 Astra
OpenAI · None effort
82.9%IndependentEpoch AI30 Aug 2026
GPT-5.6 Sol
OpenAI · Promax effort
80.5%IndependentEpoch AI9 Jul 2026
GPT-5.5 Pro
OpenAI · Xhigh effort
78.0%IndependentEpoch AI12 Jun 2026
Claude Opus 5
Anthropic · Max effort
73.2%IndependentEpoch AI24 Jul 2026
GPT-5.5
OpenAI · Xhigh effort
72.5%IndependentEpoch AI11 Jun 2026
GPT-5.6 Terra
OpenAI · Max effort
70.7%IndependentEpoch AI9 Jul 2026
GPT-5.6 Luna
OpenAI · Max effort
61.0%IndependentEpoch AI9 Jul 2026
GPT-5.4 Pro
OpenAI · Xhigh effort
58.5%IndependentEpoch AI13 Jun 2026
Claude Opus 4.8
Anthropic · Max effort
56.1%IndependentEpoch AI10 Jun 2026
GPT-5.4
OpenAI · Xhigh effort
49.0%IndependentEpoch AI11 Jun 2026
Muse Spark 1.3
Meta · Max effort
46.3%IndependentEpoch AI18 Sep 2026
Qwen3.8-Max
Alibaba Group · Xhigh effort
46.3%IndependentEpoch AI4 Aug 2026
GPT-5.2 Pro
OpenAI · Xhigh effort
46.0%IndependentEpoch AI13 Jun 2026
Muse Spark 1.3
Meta · Xhigh effort
41.5%IndependentEpoch AI16 Sep 2026
Kimi K3
Moonshot AI · Max effort
39.0%IndependentEpoch AI17 Jul 2026
Gemini 3.7 Flash
Google DeepMind · High effort
36.6%IndependentEpoch AI14 Aug 2026
Qwen3.7-Max34.1%IndependentEpoch AI13 Jun 2026
Qwen3.8-Max
Alibaba Group · Xhigh effort
34.1%IndependentEpoch AI2 Sep 2026
Claude Opus 4.7
Anthropic · Max effort
31.7%IndependentEpoch AI10 Jun 2026
Grok 4.6
SpaceXAI (xAI) · Xhigh effort
31.7%IndependentEpoch AI14 Aug 2026
GPT-5.2
OpenAI · Xhigh effort
31.7%IndependentEpoch AI11 Jun 2026
Claude Sonnet 5
Anthropic · Max effort
29.3%IndependentEpoch AI30 Jun 2026
GLM-5.2
Z.ai (Zhipu AI) · Max effort
29.3%IndependentEpoch AI19 Jun 2026
GLM-5.3
Z.ai (Zhipu AI) · Max effort
29.3%IndependentEpoch AI25 Aug 2026
Claude Opus 4.6
Anthropic · Max effort
26.8%IndependentEpoch AI11 Jun 2026
DeepSeek-V4-Pro
DeepSeek · Max effort
26.8%IndependentEpoch AI19 Aug 2026
Gemini 3.1 Pro26.8%IndependentEpoch AI11 Jun 2026
Gemini 3.5 Flash
Google DeepMind · High effort
26.8%IndependentEpoch AI10 Jun 2026
Kimi K2.625.6%IndependentEpoch AI10 Jun 2026
DeepSeek-V4-Flash
DeepSeek · Max effort
24.4%IndependentEpoch AI2 Aug 2026
Grok 4.5
SpaceXAI (xAI) · High effort
24.4%IndependentEpoch AI9 Jul 2026
Gemini 3.6 Flash
Google DeepMind · High effort
22.0%IndependentEpoch AI2 Aug 2026
Gemini 3.8 Flash
Google DeepMind · High effort
22.0%IndependentEpoch AI2 Sep 2026
GPT-5
OpenAI · High effort
22.0%IndependentEpoch AI11 Jun 2026
GPT-5 Pro
OpenAI · High effort
19.5%IndependentEpoch AI12 Jun 2026
Gemini 3 Flash17.1%IndependentEpoch AI11 Jun 2026
GLM-5.3-Flash
Z.ai (Zhipu AI) · Max effort
17.1%IndependentEpoch AI27 Aug 2026
Grok 4.2017.1%IndependentEpoch AI13 Jul 2026
Inkling Small
Thinking Machines Lab · Xhigh effort
17.1%IndependentEpoch AI14 Aug 2026
Grok 4.3
SpaceXAI (xAI) · High effort
14.6%IndependentEpoch AI17 Jun 2026
GPT-5 mini
OpenAI · High effort
12.2%IndependentEpoch AI12 Jun 2026
GPT-5.4 nano
OpenAI · High effort
12.2%IndependentEpoch AI12 Jun 2026
Kimi K2.7 Code12.2%IndependentEpoch AI13 Jun 2026
GPT-5.4 mini
OpenAI · Xhigh effort
9.8%IndependentEpoch AI12 Jun 2026
Claude Opus 4.5
Anthropic · 32K effort
4.9%IndependentEpoch AI11 Jun 2026
Inkling
Thinking Machines Lab · Xhigh effort
4.9%IndependentEpoch AI6 Aug 2026
o4-mini
OpenAI · High effort
4.9%IndependentEpoch AI11 Jun 2026
Claude Opus 4.1
Anthropic · 32K effort
2.4%IndependentEpoch AI11 Jun 2026
Claude Sonnet 4.5
Anthropic · 32K effort
2.4%IndependentEpoch AI11 Jun 2026
DeepSeek-V4-Pro
DeepSeek · Max effort
2.4%IndependentEpoch AI17 Jun 2026
GPT-5 nano
OpenAI · High effort
2.4%IndependentEpoch AI12 Jun 2026
Gemini 2.5 Pro0.0%IndependentEpoch AI11 Jun 2026
Gemini 3.5 Flash-Lite
Google DeepMind · High effort
0.0%IndependentEpoch AI2 Aug 2026
o3-mini
OpenAI · High effort
0.0%IndependentEpoch AI11 Jun 2026
  • Epoch AI, 'Capabilities & benchmarking', epoch.ai/benchmarks, CC BY 4.0.

Sources