LongCoT - Logic: leaderboard

Long-horizon chain-of-thought reasoning on the medium and hard logic problems of LongCoT (2026), expert-templated problems with short inputs and verifiable answers whose solutions need tens of thousands of reasoning tokens, single attempt without tools at the highest available reasoning setting. Accuracy in percent; higher is better.

Metric: Accuracy (%) on the medium and hard logic problems of LongCoT (main set; long-horizon chain-of-thought problems with short inputs, verifiable answers and 10K to 100K+ token reasoning), single-shot, highest available reasoning setting up to the provider output limit, no tools; higher is better. Source: arxiv.org. Saturation forecast: Around April 2028. 7 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-5.2xHigh12#185
2Gemini 3 Pro1.5#145
3Claude Sonnet 4.5Thinking0.5#261
4Kimi K2 (Thinking)0.5
5DeepSeek V3.2Thinking0.3#419
6GLM-4.70#380
7Grok 4.1 FastReasoning0#427

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=longcot-logic · How It Works · Data refreshed daily, snapshot 2026-11-21.