The FrontierMath v2 private Tiers 1-3 set contains 295 expert-written problems spanning advanced undergraduate mathematics through early-career research. Released on June 12, 2026, v2 corrected 123 Tiers 1-3 problems and removed five.
Browse the latest scores, model modes, release dates, and parameter sizes for FrontierMath v2.
Data sourced primarily from official releases (GitHub, Hugging Face, papers), then benchmark leaderboards, then third-party evaluators. Learn about our data methodology
| Rank | Model | License | |||
|---|---|---|---|---|---|
 GPT-5.6 Sol Thinking Level · Max | 89.12 | 2026-06-26 | Unknown | Closed | |
GPT-5.5 Pro Thinking Level · Extra High | 87.72 | 2026-04-23 | Unknown | Closed | |
 Claude Fable 5 Thinking Level · Max | 87.02 | 2026-06-09 | Unknown | Closed | |
4 | GPT-5.6 Terra Thinking Level · Max | 85.96 | 2026-06-26 | Unknown | Closed |
5 | Claude Opus 5 Thinking Level · Max | 85.61 | 2026-07-24 | Unknown | Closed |
6 | GPT-5.5 Thinking Level · Extra High | 85.26 | 2026-04-23 | Unknown | Closed |
7 | GPT-5.4 Pro Thinking Level · Extra High | 82.46 | 2026-03-05 | Unknown | Closed |
8 | GPT-5.6 Luna Thinking Level · Max | 82.11 | 2026-06-26 | Unknown | Closed |
9 | Claude Opus 4.8 Thinking Level · Max | 80.00 | 2026-05-28 | Unknown | Closed |
10 | GPT-5.4 Thinking Level · Extra High | 78.60 | 2026-03-05 | Unknown | Closed |
11 |  Qwen3.8-Max Thinking Level · Extra High | 74.74 | 2026-08-03 | 2400B | Closed |
12 | GPT-5.2 Pro Thinking Level · Extra High | 74.00 | 2025-12-11 | Unknown | Closed |
13 |  Kimi K3 Thinking Level · Max | 72.18 | 2026-07-16 | 2800B | Free Commercial |
14 | Opus 4.7 Thinking Level · Max | 70.18 | 2026-04-16 | Unknown | Closed |
15 | GPT-5.2 Thinking Level · Extra High | 67.40 | 2025-12-11 | Unknown | Closed |
16 | Claude Opus 4.6 Thinking Level · Max | 65.96 | 2026-02-05 | Unknown | Closed |
17 | Claude Sonnet 5 Thinking Level · Max | 65.61 | 2026-06-30 | Unknown | Closed |
18 |  Gemini 3.5 Flash Thinking Level · High | 62.81 | 2026-06-20 | Unknown | Closed |
19 |  GLM-5.2 Thinking Level · Max | 59.21 | 2026-06-13 | 753.3B | Free Commercial |
20 | Gemini 3.6 Flash Thinking Level · High | 58.95 | 2026-07-21 | Unknown | Closed |
21 |  Grok 4.5 Thinking Level · High | 57.19 | 2026-07-08 | Unknown | Closed |
22 | GPT-5-Pro Thinking Level · High | 55.79 | 2025-08-07 | Unknown | Closed |
23 | GPT-5 Thinking Level · High | 55.44 | 2025-08-07 | Unknown | Closed |
24 | GPT-5.4 mini Thinking Level · Extra High | 51.23 | 2026-03-17 | Unknown | Closed |
25 | GPT-5-mini Thinking Level · High | 46.67 | 2025-08-07 | Unknown | Closed |
26 |  DeepSeek-V4-Pro Thinking Level · Max | 45.26 | 2026-08-13 | 1600B | Free Commercial |
27 | GPT-5.4 nano Thinking Level · High | 44.91 | 2026-03-17 | Unknown | Closed |
28 | Grok 4.3 Beta Thinking Level · High | 42.81 | 2026-05-17 | 500B | Closed |
29 | OpenAI o4 - mini Thinking Level · High | 36.14 | 2025-04-16 | Unknown | Closed |
30 | Opus 4.5 32K | 34.39 | 2025-11-25 | Unknown | Closed |
In Epoch AI's data observed August 12, 2026, GPT-5.6 Sol (max) scores 89.1, GPT-5.5 Pro (xhigh) 87.7, and Claude Fable 5 (max) 87.0.
