Coding on Terminal-Bench 2.0
59.3ScoreClaude Opus 4.5
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Claude Opus 4.5Agent Framework=Terminus-22026.02 | 59.3 | — | |
| Claude Opus 4.5Agent Framework=Claude Code2026.02 | 57.9 | — | |
| GLM-5Agent Framework=Terminus-22026.02 | 56.2 | 60.7 | |
| GLM-5Agent Framework=Claude Code2026.02 | 56.2 | 61.1 | |
| Gemini 3 ProAgent Framework=Terminus-22026.02 | 54.2 | — | |
| GPT-5.2 (xhigh)Agent Framework=Terminus-22026.02 | 54 | — | |
| Kimi K2.5Agent Framework=Terminus-22026.02 | 50.8 | — | |
| DeepSeek-V3.2Agent Framework=Claude Code2026.02 | 46.4 | — | |
| GLM-4.7Agent Framework=Terminus-22026.02 | 41 | — | |
| DeepSeek-V3.2Agent Framework=Terminus-22026.02 | 39.3 | — | |
| GLM-4.7Agent Framework=Claude Code2026.02 | 32.8 | — |