OpenEvals - Terminal-Bench 2.0: leaderboard
Hugging Face OpenEvals official leaderboard covering Terminal-Bench 2.0 terminal-based agent tasks.
Metric: Success Rate (%). Source: huggingface.co. Saturation forecast: Around January 2027. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 397B A17B | 52.5 | #146 |
| 2 | GLM-5 | 52.4 | #141 |
| 3 | Step 3.5 Flash | 51 | #270 |
| 4 | Qwen 3.5 122B A10B | 49.4 | #173 |
| 5 | Kimi K2.5 | 43.2 | #144 |
| 6 | Qwen 3.5 27B | 41.6 | #187 |
| 7 | Qwen 3.5 35B A3B | 40.5 | #256 |
| 8 | DeepSeek V3.2 | 39.6 | #205 |
| 9 | Qwen3 Coder Next | 36.2 | #326 |
| 10 | Kimi K2 (Thinking) | 35.7 | #243 (Kimi K2) |
| 11 | GLM-4.7 | 33.4 | #192 |
| 12 | MiniMax-M2 | 30 | #317 |
| 13 | MiniMax-M2.1 | 29.2 | #289 |
| 14 | Kimi K2 | 27.8 | #243 |
| 15 | GLM-4.6 | 24.5 | #252 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openevals-terminal-bench-2-0 · How It Works · Data refreshed daily, snapshot 2026-10-20.