Toolathlon Pass@3 (Anthropic Internal Harness): leaderboard
Toolathlon: 108 real-world tool-use tasks across 32 applications, graded by execution-based checkers, run on Anthropic's internal harness with adaptive thinking at max effort. Pass@3 is the percentage of tasks solved in at least one of three trials (higher is better). From Table 8.11.5.A of the Claude Sonnet 5 System Card (June 2026), original task set.
Metric: Pass@3 (%). Source: www-cdn.anthropic.com. Saturation forecast: Around October 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Fable 5 (Max) | 68.5 | #20 (Claude Fable 5) |
| 2 | Claude Opus 4.8 (Max) | 67.6 | #39 (Claude Opus 4.8) |
| 3 | Claude Opus 4.6 (Max) | 66.7 | #67 (Claude Opus 4.6) |
| 4 | Claude Opus 4.7 (Max) | 66.7 | #51 (Claude Opus 4.7) |
| 5 | Claude Sonnet 5 (Max) | 63 | #65 (Claude Sonnet 5) |
| 6 | Claude Sonnet 4.6 (Max) | 60.2 | #94 (Claude Sonnet 4.6) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=toolathlon-pass-3-anthropic-internal-harness · How It Works · Data refreshed daily, snapshot 2026-11-05.