ITPEval: leaderboard

Metric: Pass@1 (%; share of all 4,680 directed statement-translation pairs whose single translated statement verifies; zero-shot direct translation at temperature 0: given a source file in Lean 4, Isabelle, Rocq or HOL Light and a target prover, emit raw target code, which must verify under the target prover). Source: arxiv.org. Saturation forecast: Around 2028. 5 models tracked.

Top models

#ModelScore
1GPT-5.529.1
2DeepSeek V4 Pro27.1
3Gemini 3.1 Pro (Preview)14
4Qwen 3 235B A22B9.7
5Claude Sonnet 4.64

Interactive version: theaggregate.ai/benchmark?slug=itpeval · How It Works · Data refreshed daily, snapshot 2026-09-29.