Long-horizon agentic tasks on HLE Full
45.8Pass@1Gemini-3.0-Pro
Evaluation results
| Method | Links | |
|---|---|---|
| Gemini-3.0-ProContext Management=w/o CM2026.03 | 45.8 | |
| Claude-4.5-OpusContext Management=w/o CM2026.03 | 43.4 | |
| GPT-5.1 HighContext Management=w/o CM2026.03 | 42.7 | |
| DeepSeek-v3.2Context Management=Baseline (w/o CM)2026.03 | 40.8 | |
| Tongyi-DR-30B-A3BContext Management=Baseline (w/o CM)2026.03 | 32.9 | |
| AgentFounder-30B-A3BContext Management=-2026.03 | 31.5 | |
| MiroThinker-v1.5-30B-A3BContext Management=-2026.03 | 31 | |
| IterResearch-30B-A3BContext Management=-2026.03 | 28.8 | |
| OpenAI DeepResearchContext Management=-2026.03 | 26.6 | |
| ASearcher-Web-32BContext Management=-2026.03 | 12.5 |