A graduate-level, Google-proof question-answering benchmark designed to evaluate expert knowledge and rigorous reasoning.
Browse the latest scores, model modes, release dates, and parameter sizes for GPQA.
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 | |||
|---|---|---|---|---|---|
 Qwen3.6-35B-A3B Thinking Enabled | 86.00 | 2026-04-16 | 35B | Free Commercial | |
 GPT-Live-1 Thinking Level · High | 84.20 | 2026-07-08 | Unknown | Closed | |
 DeepSeek-V3-0324 Standard Mode | 68.40 | 2025-03-24 | 671B | Free Commercial | |
4 |  Pangu Embedded Standard Mode | 68.00 | 2025-06-30 | 7B | Free Commercial |
5 | Qwen3-8B Standard Mode | 62.00 | 2025-04-28 | 8B | Free Commercial |
6 | DeepSeek-V3 Standard Mode | 59.10 | 2024-12-26 | 681B | Free Commercial |
7 |  GLM-4-9B-Chat Standard Mode | 58.50 | 2024-06-05 | 9B | Free Commercial |
8 |  Hunyuan-A13B-Instruct Standard Mode | 49.12 | 2025-06-27 | 80B | Free Commercial |
9 |  Mistral-Small-3.1-24B-Instruct-2503 Standard Mode | 44.42 | 2025-03-17 | 24B | Free Commercial |
10 | Mistral-Small-3.2 Standard Mode | 44.22 | 2025-06-20 | 24B | Free Commercial |
11 | Qwen3-Next Standard Mode | 43.43 | 2025-09-11 | 80B | Free Commercial |
12 | GPT-4o mini Standard Mode | 40.20 | 2024-07-18 | Unknown | Closed |
13 |  Claude 3.5 Haiku Standard Mode | 37.50 | 2024-10-22 | Unknown | Closed |
14 |  Gemma 3 - 27B (IT) Standard Mode | 36.83 | 2025-03-12 | 27B | Free Commercial |
15 |  C4AI Aya Vision 32B Standard Mode | 34.38 | 2025-03-04 | 32B | Non-Commercial |
