ArXivMath August 2026 Tools (Anthropic Internal Harness): leaderboard
Accuracy on MathArena's August 2026 set of 57 research-mathematics problems, averaged over four attempts per problem in Anthropic's internal harness without safeguard classifiers. Code execution sandbox without internet access. Distinct from MathArena's two-attempt vendor-harness leaderboard and earlier monthly sets.
Metric: Accuracy (%). Source: www-cdn.anthropic.com. Saturation forecast: Estimated already saturated. 16 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 (Max) | 96.9 | #3 (Claude Opus 5.5) |
| 2 | Claude Sonnet 5.5 (Max) | 95.2 | #10 (Claude Sonnet 5.5) |
| 3 | Claude Opus 5.5 (xHigh) | 94.7 | #3 (Claude Opus 5.5) |
| 4 | Claude Fable 5.1 (xHigh) | 93 | #9 (Claude Fable 5.1) |
| 5 | Claude Fable 5.1 (Max) | 92.1 | #9 (Claude Fable 5.1) |
| 6 | Claude Opus 5 (Max) | 90.4 | #12 (Claude Opus 5) |
| 7 | Claude Opus 5 (xHigh) | 90.4 | #12 (Claude Opus 5) |
| 8 | Claude Opus 5 (High) | 88.6 | #12 (Claude Opus 5) |
| 9 | Claude Fable 5.1 (High) | 87.7 | #9 (Claude Fable 5.1) |
| 10 | Claude Opus 5.5 (High) | 79.4 | #3 (Claude Opus 5.5) |
| 11 | Claude Opus 5 (Medium) | 78.1 | #12 (Claude Opus 5) |
| 12 | Claude Fable 5.1 (Medium) | 75.9 | #9 (Claude Fable 5.1) |
| 13 | Claude Opus 5.5 (Medium) | 67.1 | #3 (Claude Opus 5.5) |
| 14 | Claude Fable 5.1 (Low) | 61.4 | #9 (Claude Fable 5.1) |
| 15 | Claude Opus 5 (Low) | 53.1 | #12 (Claude Opus 5) |
No result here: #5 GPT-6 Astra, #7 Gemini 4 Argon, #8 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=arxivmath-august-2026-tools-anthropic-internal-harness · How It Works · Data refreshed daily, snapshot 2026-10-16.