Coding on Terminal-Bench 2.0 (Success Rate and Token Usage)
65.2Success Rate (SR)HarnessBridge
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| HarnessBridgeGenerator Model=Claude-Opus-4.72026.06 | 65.2 | 0.19 | |
| Terminus 2Generator Model=Claude-Opus-4.72026.06 | 64 | 0.26 | |
| Terminus 2Generator Model=DeepSeek-V4-Pro2026.06 | 57.3 | 1.02 | |
| HarnessBridgeGenerator Model=DeepSeek-V4-Pro2026.06 | 57.3 | 0.94 | |
| Terminus 2Generator Model=GPT-5.42026.06 | 53.9 | 9.41 | |
| HarnessBridgeGenerator Model=GPT-5.42026.06 | 53.9 | 0.99 | |
| HarnessBridgeGenerator Model=DeepSeek-V4-Flash2026.06 | 53.9 | 1.22 | |
| Terminus 2Generator Model=DeepSeek-V4-Flash2026.06 | 49.4 | 2.18 | |
| HarnessBridgeGenerator Model=GPT-5.4-Nano2026.06 | 22.5 | 0.91 | |
| Terminus 2Generator Model=GPT-5.4-Nano2026.06 | 18 | 9.8 |