Textual long-horizon complex reasoning on HLE Text
62ScoreMiroThinker-1.7
Evaluation results
| Method | Links | |
|---|---|---|
| MiroThinker-1.7Size=235B, Train=Mid/SFT/RL2026.06 | 62 | |
| MiroThinker-1.7-miniSize=30B, Train=Mid/SFT/RL2026.06 | 57.2 | |
| Doubao-2.0-ProSize=-, Train=-2026.06 | 54.2 | |
| Kimi-K2.5Size=1T, Train=-2026.06 | 52.2 | |
| Gemini-3.1-Pro-PreviewSize=-, Train=-2026.06 | 51.4 | |
| GLM-5Size=744B, Train=-2026.06 | 50.4 | |
| Claude-4.6-Sonnet-ThinkingSize=-, Train=-2026.06 | 49 | |
| Qwen3.5-397BSize=397B, Train=-2026.06 | 48.3 | |
| GPT-5.2Size=-, Train=-2026.06 | 45.5 | |
| MiniMax-M2.7Size=230B, Train=-2026.06 | 43.8 | |
| DeepSeek-V3.2Size=671B, Train=-2026.06 | 40.8 | |
| Tongyi-DeepResearchSize=30B, Train=Mid/SFT/RL2026.06 | 32.9 | |
| S1-DeepResearchSize=32B, Train=SFT2026.06 | 30.3 | |
| Qwen3-235BSize=235B, Train=-2026.06 | 11.8 | |
| Qwen3-32BSize=32B, Train=-2026.06 | 9.3 |