| # | Model | Score | Relative | Setting | Released | Run | Evidence |
|---|---|---|---|---|---|---|---|
| 1 | Kimi K2.5 Open source MoonshotAI | 86.1% | — | 27 Jan 2026 | — | Source ↗ | |
| 2 | DeepSeek V3.2 Exp Open source DeepSeek | 83.3% | high | 29 Sep 2025 | — | Source ↗ | |
| 3 | QwQ-32B Open source Alibaba | 83.3% | — | 5 Mar 2025 | — | Source ↗ | |
| 4 | Qwen3-235B-A22B-Thinking (Jul 2025) Open source Alibaba | 75.0% | — | 25 Jul 2025 | — | Source ↗ | |
| 5 | DeepSeek-R1 (May 2025) Open source DeepSeek | 75.0% | — | 28 May 2025 | — | Source ↗ | |
| 6 | Qwen3 32B Open source Qwen | 74.2% | — | 29 Apr 2025 | — | Source ↗ | |
| 7 | MiniMax-M1-80k Open source MiniMax | 69.4% | — | 13 Jun 2025 | — | Source ↗ | |
| 8 | R1 Open source DeepSeek | 69.4% | — | 20 Jan 2025 | — | Source ↗ | |
| 9 | Qwen3 235B A22B Open source Qwen | 67.7% | — | 28 Apr 2025 | — | Source ↗ | |
| 10 | Kimi K2 Instruct (0905) Open source Moonshot | 66.7% | — | 5 Sep 2025 | — | Source ↗ | |
| 11 | Qwen3 14B Open source Qwen | 62.5% | — | 29 Apr 2025 | — | Source ↗ | |
| 12 | Qwen3 8B Open source Qwen | 62.1% | — | 28 Apr 2025 | — | Source ↗ | |
| 13 | Kimi K2 Instruct Open source Moonshot | 61.1% | — | 12 Jul 2025 | — | Source ↗ | |
| 14 | Qwen3 Next 80B A3B Instruct Open source Qwen | 55.6% | — | 11 Sep 2025 | — | Source ↗ | |
| 15 | Qwen3-235B-A22B-Instruct (Jul 2025) Open source Alibaba | 52.9% | — | 25 Jul 2025 | — | Source ↗ | |
| 16 | DeepSeek V3.1 Open source DeepSeek | 52.8% | — | 21 Aug 2025 | — | Source ↗ | |
| 17 | DeepSeek-V3 (Mar 2025) Open source DeepSeek | 50.0% | — | 24 Mar 2025 | — | Source ↗ | |
| 18 | Llama 4 Maverick Open source Meta | 46.2% | — | 6 Apr 2025 | — | Source ↗ | |
| 19 | gpt-oss-120b Open source OpenAI | 44.4% | high | 5 Aug 2025 | — | Source ↗ | |
| 20 | Qwen3 30B A3B Open source Qwen | 40.6% | — | 29 Apr 2025 | — | Source ↗ | |
| 21 | Llama 4 Scout Open source Meta | 36.0% | — | 5 Apr 2025 | — | Source ↗ | |
| 22 | Gemma 3 27B Open source Google | 33.3% | — | 12 Mar 2025 | — | Source ↗ | |
| 23 | Llama 3.3 70B Open source Meta AI | 33.3% | — | 6 Dec 2024 | — | Source ↗ |
Caveats: settings (reasoning effort, agent scaffold) differ between rows and materially affect scores; the best reported setting per model is shown. Source: Epoch AI, CC BY 4.0.