Benchmark · Reasoning and knowledge

FrontierMath (Tiers 1-3)

Original research-level mathematics problems written and checked by mathematicians, answered by a Python function the model submits after reasoning with a Python tool. Epoch AI made the set and keeps most of it private; version 2 of June 2026 fixed errors in 42% of problems.

Score
% of problems solved, higher is better
Run by
Epoch AI
Status
Current, newest result 19 November 2026
Updated
as models are released
Ceiling
The best result is within five points of a perfect score, so the test no longer tells the top models apart
Data
the published file, CC BY 4.0, read 20 November 2026
Method
epoch.ai/benchmarks/frontiermath-tiers-1-3-v2

Results: Tiers 1-3 v2 (2026-06-12)

36 models we list, each at its best setting, best first.

#ModelScoreRangeWho ran itSettingDate
1Claude Fable 5.2 A95.1%Independenteffort max2026-11-06source
2GPT-6.1 Sol AA93.7%Independenteffort max2026-09-29source
2GPT-6 Astra A93.7%Independenteffort max2026-08-30source
4GPT-6.2 Astra AA92.6%Independenteffort max2026-10-27source
5Gemini 4 Neon B91.2%Independenteffort xhigh2026-11-19source
5Claude Opus 5.5 A91.2%Independenteffort max2026-09-22source
7GPT-6.1 Terra AA90.9%Independenteffort max2026-10-26source
8Gemini 4 Carbon B90.5%Independenteffort high2026-11-02source
9Claude Fable 5.1 A90.2%Independenteffort max2026-09-01source
10GPT-6 Sol A89.8%Independenteffort max2026-09-22source
11Claude Sonnet 5.5 A88.8%Independenteffort max2026-09-29source
12Gemini 4 Hydrogen B88.1%Independenteffort xhigh2026-11-13source
13Gemini 4 Argon B86.0%Independenteffort high2026-10-13source
14Muse Wonder D82.1%Independenteffort max2026-10-12source
15GPT-6 Luna A79.0%Independenteffort max2026-09-22source
16Kimi K3.1 Pro D78.2%Independenteffort max2026-11-06source
17Claude Haiku 5.5 A75.1%Independenteffort max2026-10-07source
18Muse Spark 1.3 D74.4%Independenteffort xhigh2026-09-16source
19Kimi K3 B72.2%Independenteffort max2026-07-17source
20DeepSeek V4.1 Flash C71.2%Independenteffort max2026-10-19source
21Muse Spark Open C69.8%Independenteffort xhigh2026-10-15source
22GLM-5.3 B68.8%Independenteffort max2026-08-25source
23Gemini 3.8 Flash AA68.4%Independenteffort high2026-09-02source
24Gemini 3.1 Pro (preview) AA59.6%Independent2026-06-11source
25DeepSeek-V4-Flash-0731 C57.5%Independenteffort max2026-08-02source
26GLM-5.3-Flash B55.8%Independenteffort max2026-08-27source
27Grok 4.7 BB53.0%Independenteffort xhigh2026-09-22source
28Inkling-Small B46.3%Independenteffort xhigh2026-08-14source
29DeepSeek V4 Pro D45.3%Independenteffort max2026-06-17source
30Grok 4.3 BB42.8%Independenteffort high2026-06-17source
31Qwen3.6-27B B35.1%Independent2026-08-28source
32Inkling B33.3%Independenteffort xhigh2026-08-06source
33Qwen3.5-397B-A17B B31.2%Independenteffort none2026-08-30source
34Gemini 3.1 Flash-Lite B27.7%Independenteffort high2026-08-30source
35Gemini 3.5 Flash-Lite AA26.0%Independenteffort high2026-08-02source
36Qwen3.6-35B-A3B B20.4%Independenteffort none2026-08-30source
Every setting, 43 rows
ModelAs the leaderboard names itScoreWho ran itSettingDate
Claude Fable 5.2claude-fable-5-295.1%Independenteffort max2026-11-06source
GPT-6.1 Solgpt-6.1-sol93.7%Independenteffort max2026-09-29source
GPT-6 Astragpt-6-astra93.7%Independenteffort max2026-08-30source
GPT-6.2 Astragpt-6.2-astra92.6%Independenteffort max2026-10-27source
Gemini 4 Neongemini-4-neon91.2%Independenteffort xhigh2026-11-19source
Claude Opus 5.5claude-opus-5-591.2%Independenteffort max2026-09-22source
GPT-6.1 Terragpt-6.1-terra90.9%Independenteffort max2026-10-26source
Gemini 4 Carbongemini-4-carbon90.5%Independenteffort high2026-11-02source
Claude Fable 5.1claude-fable-5-190.2%Independenteffort max2026-09-01source
GPT-6 Solgpt-6-sol89.8%Independenteffort max2026-09-22source
Claude Sonnet 5.5claude-sonnet-5-588.8%Independenteffort max2026-09-29source
Gemini 4 Hydrogengemini-4-hydrogen88.1%Independenteffort xhigh2026-11-13source
Gemini 4 Argongemini-4-argon86.0%Independenteffort high2026-10-13source
Muse Wondermuse-wonder-1.082.1%Independenteffort max2026-10-12source
GPT-6 Lunagpt-6-luna79.0%Independenteffort max2026-09-22source
Kimi K3.1 Prokimi-k3.1-pro-together78.2%Independenteffort max2026-11-06source
Claude Haiku 5.5claude-haiku-5-575.1%Independenteffort max2026-10-07source
Muse Spark 1.3muse-spark-1.374.4%Independenteffort xhigh2026-09-16source
Muse Spark 1.3muse-spark-1.374.0%Independenteffort max2026-09-18source
Kimi K3kimi-k372.2%Independenteffort max2026-07-17source
DeepSeek V4.1 Flashdeepseek-v4.1-flash-101771.2%Independenteffort max2026-10-19source
Muse Spark OpenMuse-Spark-Open-816B-A42B69.8%Independenteffort xhigh2026-10-15source
GLM-5.3glm-5.368.8%Independenteffort max2026-08-25source
Gemini 3.8 Flashgemini-3.8-flash68.4%Independenteffort high2026-09-02source
DeepSeek V4.1 Flashdeepseek-v4.1-flash67.4%Independenteffort max2026-10-08source
Gemini 3.1 Pro (preview)gemini-3.1-pro-preview59.6%Independent2026-06-11source
DeepSeek-V4-Flash-0731deepseek-v4-flash-073157.5%Independenteffort max2026-08-02source
GLM-5.3-Flashglm-5.3-flash55.8%Independenteffort max2026-08-27source
Grok 4.7grok-4.753.0%Independenteffort xhigh2026-09-22source
Inkling-Smallinkling-small46.3%Independenteffort xhigh2026-08-14source
DeepSeek V4 Prodeepseek-v4-pro45.3%Independenteffort max2026-06-17source
Grok 4.3grok-4.342.8%Independenteffort high2026-06-17source
Qwen3.6-27Bqwen3.6-27b35.1%Independent2026-08-28source
Qwen3.6-27Bqwen3.6-27b34.0%Independenteffort none2026-08-29source
InklingInkling33.3%Independenteffort xhigh2026-08-06source
Qwen3.5-397B-A17Bqwen3.5-397b-a17b31.2%Independenteffort none2026-08-30source
Qwen3.5-397B-A17Bqwen3.5-397b-a17b29.5%Independent2026-08-28source
Gemini 3.1 Flash-Litegemini-3.1-flash-lite27.7%Independenteffort high2026-08-30source
Gemini 3.5 Flash-Litegemini-3.5-flash-lite26.0%Independenteffort high2026-08-02source
Gemini 3.1 Flash-Litegemini-3.1-flash-lite22.5%Independenteffort low2026-08-30source
Gemini 3.1 Flash-Litegemini-3.1-flash-lite21.4%Independenteffort minimal2026-08-28source
Qwen3.6-35B-A3Bqwen3.6-35b-a3b20.4%Independenteffort none2026-08-30source
Qwen3.6-35B-A3Bqwen3.6-35b-a3b17.5%Independent2026-08-29source

Results: Tiers 1-3 (2025-02-28)

1 model we list, each at its best setting, best first.

#ModelScoreRangeWho ran itSettingDate
1Gemini 3.1 Pro (preview) AA36.9%Independent2026-02-19source

A rank is among the models we list with a result, not a place on the whole leaderboard, and a model offered by several hosts is shown once, on its own listing. These results don't move our grades. Data: Epoch AI, Capabilities and benchmarking (CC BY 4.0), CC BY 4.0.

For companies

Do agents find, use and choose your tools?

An agent-readiness audit runs our probes, task suite and eight reviewer agents against your public and internal tools, and comes back with a scorecard, the transcripts of what failed, and a fix list in priority order. From $2,500, re-run included. We never take payment to move a rank. We do help companies earn one.