Terminal-Bench 2.1 (Best Reported Harness): leaderboard
Provider-reported Terminal-Bench 2.1 comparison using each model's best reported agent harness rather than a single shared harness.
Metric: Score (%). Source: z.ai. Source: benchmarklist.com. Saturation forecast: Around January 2027. 12 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5 (xHigh) | 83.82 | #16 (Claude Fable 5) |
| 2 | GPT-5.5 (xHigh) | 83.15 | #31 (GPT-5.5) |
| 3 | Grok 4.5 (High) | 79.33 | #48 (Grok 4.5) |
| 4 | Claude Opus 4.8 (High) | 78.88 | #37 (Claude Opus 4.8) |
| 5 | GPT-5.6 Terra (Max) | 78.43 | #41 (GPT-5.6 Terra) |
| 6 | Muse Spark 1.1 (xHigh) | 76.18 | #52 (Muse Spark 1.1) |
| 7 | GPT-5.6 Luna (Max) | 75.73 | #68 (GPT-5.6 Luna) |
| 8 | Claude Sonnet 5 (High) | 74.61 | #57 (Claude Sonnet 5) |
| 9 | Gemini 3 Pro (Preview) (High) | 73.93 | #66 (Gemini 3 Pro (Preview)) |
| 10 | Claude Opus 4.7 (Max) | 68.9 | #49 (Claude Opus 4.7) |
| 11 | Gemini 3.1 Pro (Preview) (High) | 65.84 | #58 (Gemini 3.1 Pro (Preview)) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=terminal-bench-2-1-best-reported-harness · How It Works · Data refreshed daily, snapshot 2026-10-20.