ClosureBench (Chain-of-Thought) - L3 Composition: leaderboard

Metric: Accuracy (%; lenient scoring of one JSON answer per question; L3 compositional queries (three to five steps, 350 instances); held-out split of 1,300 generated graph questions over graph sizes 4-10, temperature 0; chain-of-thought mode: reason step by step, then answer). Source: arxiv.org. Saturation forecast: Around December 2026. 4 models tracked.

Top models

#ModelScore
1GPT-4.1 Mini74.9
2GPT-4.174.6
3Claude Sonnet 465.1
4Gemini 2.5 Flash63.1

Interactive version: theaggregate.ai/benchmark?slug=closurebench-chain-of-thought-l3-composition · How It Works · Data refreshed daily, snapshot 2026-09-26.