LemmaBench (April 2026, Step-by-Step, GPT-5.5 Judge): leaderboard
Metric: Step-by-step pass rate (%): share of the April 2026 batch's 120 self-contained research-level lemmas whose generated proof has every step judged correct by a GPT-5.5 judge giving a binary verdict per step (stricter than whole-proof judging); higher is better. Source: arxiv.org. Saturation forecast: Around July 2028. 3 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.5 | 20.8 | #26 |
| 2 | Gemini 3.1 Pro (Preview) | 8.3 | #54 |
| 3 | GPT-5.4 Mini | 3.3 | #202 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=lemmabench-april-2026-step-by-step-gpt-5-5-judge · How It Works · Data refreshed daily, snapshot 2026-10-11.