Agentic · terminal
Terminal-Bench 2.1 leaderboard
As of 2026-10-07, the highest measured score on Terminal-Bench 2.1 is 91.4% by Claude Fable 5.1 (max with fallback). 17 more models have estimated scores, calibrated from the benchmarks they were measured on.
Measured scores: artificial-analysis.
| # | Model | Score | Source |
|---|---|---|---|
| 1 | Claude Fable 5.1 (max with fallback) | 91.4% | measured |
| 2 | Claude Fable 5.1 (xhigh with fallback) | 91.0% | measured |
| 3 | Claude Fable 5.1 (high with fallback) | 89.9% | measured |
| 4 | GPT-6 Astra (high) | 89.9% | measured |
| 5 | GPT-6 Astra (medium) | 89.5% | measured |
| 6 | GPT-5.6 Sol (xhigh) | 89.5% | measured |
| 7 | Claude Opus 5.5 (medium with fallback) | 89.4% | estimated ± 7.0 pp, medium confidence |
| 8 | Claude Sonnet 5.5 (high with fallback) | 89.1% | estimated ± 7.0 pp, medium confidence |
| 9 | Qwen3.8 Max (0902) | 88.8% | measured |
| 10 | GPT-6 Astra (max) | 88.4% | measured |
| 11 | Claude Sonnet 5.5 (max with fallback) | 88.3% | estimated ± 4.0 pp, medium confidence |
| 12 | Claude Opus 5.5 (max with fallback) | 88.2% | estimated ± 4.0 pp, medium confidence |
| 13 | Claude Opus 5.5 (xhigh with fallback) | 88.2% | estimated ± 4.0 pp, medium confidence |
| 14 | GPT-6 Astra (xhigh) | 88.2% | estimated ± 4.0 pp, medium confidence |
| 15 | Gemini 4 Argon (high) | 88.2% | estimated ± 4.0 pp, high confidence |
| 16 | Claude Sonnet 5.5 (xhigh with fallback) | 88.2% | estimated ± 4.0 pp, high confidence |
| 17 | Claude Opus 5.5 (high with fallback) | 88.2% | estimated ± 4.0 pp, high confidence |
| 18 | GPT-6.1 Sol (max) | 88.2% | estimated ± 4.0 pp, high confidence |
| 19 | GPT-6 Sol (max) | 87.8% | estimated ± 4.0 pp, high confidence |
| 20 | Gemini 3.8 Flash (high) | 87.6% | measured |
| 21 | MiMo-V2.6-Pro | 87.4% | estimated ± 4.0 pp, high confidence |
| 22 | Step 5 Preview | 87.4% | estimated ± 4.0 pp, high confidence |
| 23 | DeepSeek V4.1 Flash (max) | 86.9% | estimated ± 4.0 pp, high confidence |
| 24 | Mistral Large 4 Preview | 86.9% | estimated ± 4.0 pp, high confidence |
| 25 | Grok 4.7 (xhigh) | 86.8% | estimated ± 4.0 pp, high confidence |
| 26 | Kimi K3 (max) | 85.0% | measured |
| 27 | GPT-6 Luna (max) | 84.3% | estimated ± 4.0 pp, high confidence |
| 28 | Muse Spark 1.3 (max) | 84.3% | measured |
| 29 | GLM-5.3-Flash | 84.3% | measured |
| 30 | GLM-5.3 (max) | 83.9% | measured |
| 31 | Qwen3.8 27B (xhigh) | 79.8% | measured |
| 32 | K2 Horizon 375B A23B | 71.9% | measured |
| 33 | MiniMax-M3 | 65.2% | measured |
| 34 | Inkling (xhigh) | 55.1% | measured |
| 35 | Nemotron 3 Ultra | 53.9% | measured |
| 36 | Muse Glimmer (high) | 51.7% | measured |
All leaderboards
Agentic · terminal
Agentic · tools
- AA-AnalystAgent
- APEX-Agents
- AutomationBench
- EnterpriseOps-Gym
- Harvey LAB
- ITBench SRE
- τ³-Bench Banking
- τ²-Bench Telecom