Terminal-Bench 4.0 (Claude Code): leaderboard
Terminal-Bench4.0 reported resolution rate (%) with Claude Code. Each primary row retains explicit source reasoning effort, success/trial counts and confidence interval; different efforts remain separate. Literal source effort none is not inferred to mean disabled reasoning.
Metric: Resolution Rate (%). Source: www.tbench.ai. Saturation forecast: Around July 2027. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 (Max) | 64.85 | #3 (Claude Opus 5.5) |
| 2 | Claude Sonnet 5.5 (Max) | 61.82 | #11 (Claude Sonnet 5.5) |
| 3 | Claude Fable 5.1 (Max) | 57.88 | #9 (Claude Fable 5.1) |
| 4 | Claude Fable 5.1 (xHigh) | 57.88 | #9 (Claude Fable 5.1) |
| 5 | Claude Fable 5.1 (High) | 54.55 | #9 (Claude Fable 5.1) |
| 6 | Claude Opus 5 (xHigh) | 53.94 | #12 (Claude Opus 5) |
| 7 | Claude Fable 5.1 (Medium) | 53.94 | #9 (Claude Fable 5.1) |
| 8 | Claude Opus 5 (Max) | 51.82 | #12 (Claude Opus 5) |
| 9 | Claude Opus 5 (High) | 50.3 | #12 (Claude Opus 5) |
| 10 | Claude Opus 5 (Medium) | 44.85 | #12 (Claude Opus 5) |
| 11 | Claude Fable 5 (Max) | 44.55 | #16 (Claude Fable 5) |
| 12 | Claude Fable 5.1 (Low) | 43.33 | #9 (Claude Fable 5.1) |
| 13 | GLM-5.3 (Max) | 41.82 | #39 (GLM-5.3) |
| 14 | Claude Opus 5 (Low) | 34.85 | #12 (Claude Opus 5) |
| 15 | Claude Opus 4.8 (Max) | 23.64 | #37 (Claude Opus 4.8) |
No result here: #6 GPT-6 Astra, #7 GPT-6.1 Sol, #8 Gemini 4 Argon.
Interactive version: theaggregate.ai/benchmark?slug=terminal-bench-4-0-claude-code · How It Works · Data refreshed daily, snapshot 2026-10-20.