HyperLogic - Base (Item): leaderboard
HyperLogic (Tencent Hunyuan and Fudan, arXiv 2605.19597v2; first released as LLMEval-Logic) is a Chinese logical-reasoning set built by forward construction: problems are written in natural language first and formalized afterwards, with each answer derived by two independent formalizations and logic solvers. Unlike reverse-constructed sets such as ZebraLogic, which render a predefined formal structure into text, no formalization is fixed in advance, so "choosing a faithful model is itself part of the reasoning task". Every model runs at its highest reasoning setting (Grok 4.6 xhigh, the others max); free-form Chinese answers are graded by a semantic-equivalence LLM judge, and the score is accuracy (%) on the full set, the mean of three rollouts. This board: item-level accuracy on the 195 Base items, each a shared background with several sub-questions. Claude Opus 5's cell excludes items its API refused, so it is not read here. Higher is better. Source: https://arxiv.org/abs/2605.19597
Metric: Item accuracy (%). Source: arxiv.org. Saturation forecast: Around March 2027. 6 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.6 Sol (Max) | 77.6 | #22 (GPT-5.6 Sol) |
| 2 | Kimi K3 (Max) | 57.9 | #35 (Kimi K3) |
| 3 | Grok 4.6 (xHigh) | 52.5 | #29 (Grok 4.6) |
| 4 | DeepSeek V4 Pro (Max) | 51.8 | #112 (DeepSeek V4 Pro) |
| 5 | GLM-5.3 (Max) | 45.6 | #44 (GLM-5.3) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=hyperlogic-base-item · How It Works · Data refreshed daily, snapshot 2026-11-05.