Benchmarks / Agentic
Terminal-Bench 4.0
Resolution rate of agent+model systems on 66 professional computer-work tasks in a terminal (5 trials per task, 8-hour timeout). Rows are harness+model pairs.
Terminal-Bench 4.0
Higher is better · best effort per model
full table below
| # | Model | Effort | Score |
|---|---|---|---|
| 1 | Opus 5.5 | max | 64.8% |
| 2 | Sonnet 5.5 | max | 61.8% |
| 3 | GPT-6 Astra | max | 58.2% |
| 4 | GPT-6.1 Sol | max | 58.2% |
| 5 | Fable 5.1 | max | 57.9% |
| 6 | Opus 5 | xhigh | 53.9% |
| 7 | GPT-6 Sol | max | 49.4% |
| 8 | Fable 5 | max | 44.5% |
| 9 | GLM 5.3 | max | 41.8% |
| 10 | Grok 4.7 | xhigh | 37.6% |
| 11 | GPT-5.6 Sol | max | 37.3% |
| 12 | GLM 5.3 Flash | none | 35.8% |
| 13 | Qwen 3.8 Max | max | 27.0% |
| 14 | Opus 4.8 | max | 23.6% |
| 15 | GPT-5.6 Terra | max | 21.5% |
| 16 | Grok 4.6 | high | 20.3% |
| 17 | Gemini 3.8 Flash | high | 19.1% |
| 18 | GPT-5.6 Luna | max | 17.3% |
| 19 | GPT-6 Luna | max | 16.4% |
| 20 | Muse Spark 1.3 | xhigh | 14.6% |
| 21 | Grok 4.5 | high | 12.4% |
| 22 | Sonnet 5 | max | 12.4% |
| 23 | Gemini 3.7 Flash | high | 11.2% |