Personal Assistant Agent Performance on Claw-Eval general domain 0408
70.8Pass@3Claude Opus 4.6
Evaluation results
| Method | Links | |
|---|---|---|
| Claude Opus 4.62026.05 | 70.8 | |
| GPT 5.42026.05 | 60.2 | |
| Qwen3.5 397A17B2026.05 | 57.8 | |
| Gemini 3.1 Pro2026.05 | 55.9 | |
| GLM 5 Turbo2026.05 | 52.8 | |
| MiniMax M2.72026.05 | 49.7 | |
| MiniMax M2.52026.05 | 47.2 | |
| Kimi K2.52026.05 | 36.6 | |
| Orchard-Claw (SFT + RL)Model Size=30B-A3B (~3B active), #Tasks=0.2k2026.05 | 31.7 | |
| Qwen3-Coder-30B-A3B-InstructModel Size=30B-A3B (~3B active)2026.05 | 30.4 | |
| Nemotron-3-nano-30b-a3bModel Size=30B-A3B (~3B active)2026.05 | 26.1 | |
| Orchard-Claw (SFT)Model Size=30B-A3B (~3B active), #Tasks=0.2k2026.05 | 22.4 | |
| Qwen3-30B-A3B-ThinkingModel Size=30B-A3B (~3B active)2026.05 | 14.3 |