OckBench: leaderboard
OckBench — Occam's-razor reasoning: choosing minimal sufficient explanations; accuracy.
Metric: Accuracy (%). Source: ockbench.github.io. Saturation forecast: Around January 2027. 56 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5 (xHigh) | 95 | #12 (Claude Opus 5) |
| 2 | Claude Opus 5 (High) | 94 | #12 (Claude Opus 5) |
| 3 | Claude Opus 4.8 (xHigh) | 91.5 | #37 (Claude Opus 4.8) |
| 4 | GPT-5.5 (xHigh) | 90 | #31 (GPT-5.5) |
| 5 | Claude Opus 4.8 (Max) | 90 | #37 (Claude Opus 4.8) |
| 6 | GPT-5.5 (High) | 88.5 | #31 (GPT-5.5) |
| 7 | Gemini 3.1 Pro (Preview) (Medium) | 87.5 | #58 (Gemini 3.1 Pro (Preview)) |
| 8 | Kimi K3 (High) | 87.5 | #30 (Kimi K3) |
| 9 | GPT-5.5 (Medium) | 86 | #31 (GPT-5.5) |
| 10 | Claude Opus 4.8 (High) | 86 | #37 (Claude Opus 4.8) |
| 11 | Claude Opus 4.8 (Medium) | 86 | #37 (Claude Opus 4.8) |
| 12 | Claude Opus 5 (Medium) | 85 | #12 (Claude Opus 5) |
| 13 | Claude Opus 5 (Low) | 85 | #12 (Claude Opus 5) |
| 14 | Claude Opus 4.6 | 84.5 | #63 |
| 15 | DeepSeek V4 Pro (Max) | 84 | #102 (DeepSeek V4 Pro) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ockbench · How It Works · Data refreshed daily, snapshot 2026-10-20.