Conceptual Reasoning Index - Consistency (ACCoRD): leaderboard
The consistency component of the CRI. Separate prompts ask the model for probabilities of related claims, and 567 expert-checked constraints, such as a conjunction being no more probable than its conjuncts, measure how far the stated beliefs drift. Violations are averaged and normalized against unrelated probability gaps, then scaled so uniform random answering is 0 and perfect consistency 100, higher is better.
Metric: Chance-Corrected Score (0-100). Source: conceptualreasoning.ai. Saturation forecast: Around January 2027. 210 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 5.5 (High) | 88.42 | #3 (Claude Opus 5.5) |
| 2 | Claude Opus 5.5 | 86.45 | #3 |
| 3 | Claude Opus 5.5 (Medium) | 85.7 | #3 (Claude Opus 5.5) |
| 4 | Claude Sonnet 5.5 (xHigh) | 85.66 | #11 (Claude Sonnet 5.5) |
| 5 | Claude Sonnet 5.5 | 85.09 | #11 |
| 6 | Claude Fable 5 | 84.95 | #16 |
| 7 | Claude Opus 5.5 (Low) | 84.94 | #3 (Claude Opus 5.5) |
| 8 | Claude Opus 5.5 (xHigh) | 84.73 | #3 (Claude Opus 5.5) |
| 9 | Claude Opus 5 (Low) | 83.84 | #12 (Claude Opus 5) |
| 10 | Claude Fable 5.1 (Medium) | 83.67 | #9 (Claude Fable 5.1) |
| 11 | Claude Opus 5 (xHigh) | 83.47 | #12 (Claude Opus 5) |
| 12 | Claude Sonnet 5.5 (Medium) | 83.39 | #11 (Claude Sonnet 5.5) |
| 13 | Claude Sonnet 5.5 (Low) | 82.92 | #11 (Claude Sonnet 5.5) |
| 14 | GPT-6 Sol | 82.4 | #14 |
| 15 | Claude Fable 5 (Medium) | 82.37 | #16 (Claude Fable 5) |
No result here: #10 Muse Wonder Full.
Interactive version: theaggregate.ai/benchmark?slug=conceptual-reasoning-index-consistency-accord · How It Works · Data refreshed daily, snapshot 2026-10-20.