Terminal Task Execution on Terminal-Bench 1.0
51AccuracyClaude Sonnet 4.5
Evaluation Results
| Method | Links | |
|---|---|---|
| Claude Sonnet 4.5Model Category=Proprietary, Scaffold=Terminus 22026.04 | 51 | |
| Claude Opus 4.1Model Category=Proprietary, Scaffold=Terminus 22026.04 | 43.8 | |
| Claude Haiku 4.5Model Category=Proprietary, Scaffold=Terminus 22026.04 | 41.8 | |
| GPT-5Model Category=Proprietary, Scaffold=Terminus 22026.04 | 41.3 | |
| Claude Opus 4Model Category=Proprietary, Scaffold=Terminus 22026.04 | 39 | |
| Grok 4Model Category=Open-Source, Scaffold=Terminus 22026.04 | 39 | |
| Claude Sonnet 4Model Category=Proprietary, Scaffold=Terminus 22026.04 | 36.4 | |
| Qwen3-32B + SSModel Category=Open-Source, Scaffold=Terminus 2, Training Trajectories=SkillSynth2026.04 | 33.8 | |
| Grok 4 FastModel Category=Open-Source, Scaffold=Terminus 22026.04 | 31.3 | |
| GPT-5-MiniModel Category=Proprietary, Scaffold=Terminus 22026.04 | 30.8 | |
| Qwen3-14B + SSModel Category=Open-Source, Scaffold=Terminus 2, Training Trajectories=SkillSynth2026.04 | 22.9 | |
| Qwen3-8B + SSModel Category=Open-Source, Scaffold=Terminus 2, Training Trajectories=SkillSynth2026.04 | 17.1 |