Long-Horizon Terminal-Bench — leaderboard
Metric: Mean Score (%). Source: zli12321.github.io. 20 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Grok 4.5 | 50.5 |
| 2 | Claude Sonnet 5 | 49.7 |
| 3 | Claude Opus 4.8 | 49.2 |
| 4 | Claude Fable 5 | 48.7 |
| 5 | GPT-5.6 Sol | 45.1 |
| 6 | GPT-5.5 | 44.5 |
| 7 | MiniMax-M3 | 38.5 |
| 8 | Claude Sonnet 4.6 | 37.3 |
| 9 | Kimi K2.7 Code | 36.7 |
| 10 | GLM-5.2 | 31.6 |
| 11 | Qwen 3.6 Plus | 31.3 |
| 12 | DeepSeek V4 Pro | 30.7 |
| 13 | Qwen 3.7 Max | 29.6 |
| 14 | Hy3 | 28.8 |
| 15 | Gemini 3.1 Pro (Preview) | 27.9 |
Interactive version: aibenchmarks.dev/benchmark?slug=long-horizon-terminal-bench · How the rankings work · Data refreshed daily, snapshot 2026-07-16.