Terminal
Terminal Agent
How well agent harnesses complete real shell-based tasks.
Last updated · May 08, 2026, 12:40 UTC
| Rank | Model | Composite | Trend |
|---|---|---|---|
| 1 | opencode / DeepSeek V4 Flash opencode | 89.7 | |
| 2 | zero / DeepSeek V4 Flash zero | 88.5 | |
| 3 | claude-code / DeepSeek V4 Flash claude-code | 86.1 | |
| 4 | opencode / GLM-5.1 opencode | 85.5 | |
| 5 | zero / DeepSeek V4 Pro zero | 84.9 | |
| 6 | claude-code / DeepSeek V4 Pro claude-code | 84.2 | |
| 7 | droid / GLM-5.1 droid | 84.2 | |
| 8 | deeptide / DeepSeek V4 Pro deeptide | 83.6 | |
| 9 | zero / Claude Opus 4.5 zero | 80.9 | |
| 10 | deeptide / DeepSeek V4 Flash deeptide | 78.2 | |
| 11 | a8e / GLM-5.1 a8e | 76.4 | |
| 12 | opencode / DeepSeek V4 Pro opencode | 75.8 | |
| 13 | droid / DeepSeek V4 Pro droid | 73.9 |