VeriContest - ProofGen: leaderboard
Metric: Proof generation pass@1 (%): share of problems for which the model, given the ground-truth specification and code (spec_code2proof), writes Verus proofs that verify without modifying the specification or code, on the 946 VeriContest competitive-programming problems (690 LeetCode, 256 Codeforces) in Rust with Verus, one-shot prompting with a Verus-Bench example, temperature 0 where the API exposes it (provider default for GPT-5.5 and Claude Opus 4.7); higher is better. Source: arxiv.org. Saturation forecast: Around 2029. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 13.95 |
| 2 | Gemini 3.1 Pro (Preview) | 13.53 |
| 3 | Claude Sonnet 4.6 | 13.21 |
| 4 | Claude Opus 4.7 | 12.68 |
| 5 | GPT-5.4 Mini | 7.08 |
| 6 | Gemini 3 Flash (Preview) | 6.34 |
| 7 | DeepSeek V4 Pro | 5.6 |
| 8 | GLM-4.7 Flash | 5.5 |
| 9 | DeepSeek V4 Flash | 4.86 |
| 10 | Qwen 3.6 27B | 4.86 |
Interactive version: theaggregate.ai/benchmark?slug=vericontest-proofgen · How It Works · Data refreshed daily, snapshot 2026-10-07.