Multimodal long-horizon complex reasoning on HLE-VL
25.1ScoreGemini-3.1-Pro-Preview
Evaluation results
| Method | Links | |
|---|---|---|
| Gemini-3.1-Pro-PreviewSize=-, Train=-2026.06 | 25.1 | |
| Claude-4.6-Sonnet-ThinkingSize=-, Train=-2026.06 | 24.3 | |
| REDSearcher-MM-SFTSize=30B, Train=Mid/SFT2026.06 | 24.2 | |
| Kimi-K2.5Size=1T, Train=-2026.06 | 22.3 | |
| GPT-5.2Size=-, Train=-2026.06 | 20.8 | |
| GLM-5Size=744B, Train=-2026.06 | 19.9 | |
| Qwen3.5-397BSize=397B, Train=-2026.06 | 18.7 | |
| Doubao-2.0-ProSize=-, Train=-2026.06 | 17.5 | |
| S1-DeepResearchSize=32B, Train=SFT2026.06 | 15.2 | |
| DeepSeek-V3.2Size=671B, Train=-2026.06 | 14.9 | |
| MiniMax-M2.7Size=230B, Train=-2026.06 | 13.1 | |
| Qwen3-32BSize=32B, Train=-2026.06 | 11.4 | |
| Qwen3-235BSize=235B, Train=-2026.06 | 9.6 |