SuperGPQA — Broad grad-level QA bench (GPQA extended); folds naming variants. Distinct from GPQA and EM variant.
| # | Model | Vendor | Best score | Runs | Last seen |
|---|---|---|---|---|---|
| 1 | Qwen3.7 Max | Alibaba | 73.6 | 1 | 2026-08-23 |
| 2 | Qwen3.6 Plus | Alibaba | 71.6 | 1 | 2026-08-23 |
| 3 | Qwen3.7 Plus Preview | Alibaba | 71.4 | 1 | 2026-08-23 |
| 4 | Claude Opus 4.5 | Anthropic | 70.6 | 1 | 2026-06-15 |
| 5 | Qwen3.5 397B A17B | Alibaba | 70.4 | 2 | 2026-08-23 |
| 6 | Qwen3.5 122B A10B | Alibaba | 67.1 | 1 | 2026-08-23 |
| 7 | Qwen3.6 27B | Alibaba | 66 | 2 | 2026-08-23 |
| 8 | Gemma 4 31B | 65.7 | 2 | 2026-06-15 | |
| 9 | Qwen3.5 27B | Alibaba | 65.6 | 3 | 2026-08-23 |
| 10 | Qwen3 Max (Reasoning) | Alibaba | 65.1 | 1 | 2026-08-23 |
| 11 | Qwen3.6 35B A3B | Alibaba | 64.7 | 3 | 2026-08-23 |
| 12 | Qwen3 VL 235B A22B Reasoning | Alibaba | 64.3 | 1 | 2026-08-23 |
| 13 | Qwen3.5 35B A3B | Alibaba | 63.4 | 2 | 2026-08-23 |
| 14 | Qwen3 235B A22B Instruct 2507 | Alibaba | 62.6 | 1 | 2026-08-23 |
| 15 | Gemma 4 26B A4B | 61.4 | 1 | 2026-06-06 | |
| 16 | Qwen3 Next 80B A3B | Alibaba | 60.8 | 1 | 2026-08-23 |
| 17 | Qwen3 VL 235B A22B Instruct | Alibaba | 60.4 | 1 | 2026-08-23 |
| 18 | Qwen3 VL 32B Reasoning | Alibaba | 59 | 1 | 2026-08-23 |
| 19 | Qwen3 Next 80B A3B Instruct | Alibaba | 58.8 | 1 | 2026-08-23 |
| 20 | Qwen3.5 9B | Alibaba | 58.2 | 1 | 2026-08-23 |
| 21 | Qwen3 30B A3B | Alibaba | 57.3 | 1 | 2026-06-15 |
| 22 | Kimi K2 Instruct | Moonshot | 57.2 | 4 | 2026-08-23 |
| 23 | Claude 4 Opus | Anthropic | 56.5 | 2 | 2026-06-15 |
| 24 | Qwen3 VL 30B A3B Reasoning | Alibaba | 56.4 | 1 | 2026-08-23 |
| 25 | Claude Sonnet 4 | Anthropic | 55.7 | 2 | 2026-06-15 |
| 26 | Seed Oss 36B Instruct | ByteDance | 55.7 | 1 | 2026-06-15 |
| 27 | Qwen3 VL 32B Instruct | Alibaba | 54.6 | 1 | 2026-08-23 |
| 28 | DeepSeek-V3 | DeepSeek | 53.7 | 2 | 2026-06-15 |
| 29 | Qwen3 VL 30B A3B Instruct | Alibaba | 53.1 | 1 | 2026-08-23 |
| 30 | Qwen3.5 4B | Alibaba | 52.9 | 1 | 2026-08-23 |
| 31 | Hy3 preview-Base | Tencent | 51.6 | 2 | 2026-06-05 |
| 32 | Qwen3 VL Thinking (8B) | Alibaba | 51.2 | 1 | 2026-08-23 |
| 33 | Kimi K2 Base | Moonshot | 51.1 | 7 | 2026-08-23 |
| 34 | Qwen3 235B A22B | Alibaba | 50.2 | 3 | 2026-08-23 |
| 35 | GLM-4.5-Base | Z.ai | 49.6 | 2 | 2026-06-05 |
| 36 | Qwen3 32B | Alibaba | 49.3 | 1 | 2026-06-15 |
| 37 | Claude 3.5 Sonnet | Anthropic | 48.2 | 5 | 2026-06-05 |
| 38 | Qwen3 VL 4B (Reasoning) | Alibaba | 46.8 | 1 | 2026-08-23 |
| 39 | DeepSeek-V3-Base | DeepSeek | 46.2 | 2 | 2026-06-05 |
| 40 | OpenAI o1-mini | OpenAI | 45.2 | 5 | 2026-06-05 |
| 41 | Qwen3 VL 8B Instruct | Alibaba | 44.5 | 1 | 2026-08-23 |
| 42 | QwQ 32B Preview | Alibaba | 43.6 | 5 | 2026-06-05 |
| 43 | DeepSeek-V3.2-Exp-Base | DeepSeek | 43.6 | 4 | 2026-06-13 |
| 44 | GPT-4o | OpenAI | 42.4 | 5 | 2026-06-05 |
| 45 | DeepSeek-V3.1-Base | DeepSeek | 42.3 | 4 | 2026-06-13 |
| 46 | MiMo V2 Flash Base | Xiaomi | 41.1 | 4 | 2026-06-13 |
| 47 | Step3.5 Flash Base | StepFun | 41 | 2 | 2026-06-05 |
| 48 | R1-Distill-Qwen-14B | Alibaba | 40.6 | 5 | 2026-06-05 |
| 49 | MiMo 7B RL | Xiaomi | 40.5 | 5 | 2026-06-05 |
| 50 | Qwen3 VL 4B Instruct | Alibaba | 40.3 | 1 | 2026-08-23 |
| 51 | Qwen3.5 2B | Alibaba | 37.5 | 1 | 2026-08-23 |
| 52 | R1-Distill-Qwen-7B | Alibaba | 28.9 | 5 | 2026-06-05 |
| 53 | Qwen2.5 7B Base | Alibaba | 26.3 | 1 | 2026-05-03 |
| 54 | Qwen3.5 0.8B | Alibaba | 21.3 | 1 | 2026-08-23 |
| 55 | Qwen3 4B Base Dense | Alibaba | 20.9 | 1 | 2026-05-03 |
Best tracked score per model (default configuration; source-attributed and verification-tiered). Open a model for its full benchmark surface, provenance and pricing.