ChaosBench-Logic v2: leaderboard
Standard accuracy on binary reasoning benchmarks hides critical failure modes: prior collapse, inconsistency under paraphrase, and inability to reason about parameter-dependent dynamics. We present ChaosBench-Logic v2, a 40,886-question benchmark over 165 dynamical systems with 27 FOL predicates and 78 axiom edges, together with CARE (Calibration- and Adversarial-Robust Evaluation), a protocol tha
Metric: MCC (self-reported). Source: benchmarklist.com. Saturation forecast: Around April 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.6 | 60.1 | #90 |
| 2 | GPT-5.2 | 50.9 | #110 |
| 3 | Qwen 2.5 32B | 47.8 | #546 |
| 4 | DeepSeek V3 Chat | 46.9 | #447 |
| 5 | Gemini 2.5 Flash | 45.8 | #245 |
| 6 | GPT-4o | 45 | #343 |
| 7 | Qwen 2.5 14B | 42.6 | #725 |
| 8 | Llama 3.3 70B | 37.3 | #579 |
| 9 | Llama 3.1 8B | 24 | #1151 |
| 10 | Mistral 7B | 22.8 | #1397 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=chaosbench-logic-v2 · How It Works · Data refreshed daily, snapshot 2026-10-20.