VeriContest - CodeGen (Natural Language): leaderboard
Metric: Code generation pass@1 (%) from the natural-language problem description (nl2code): share of problems whose generated Rust code passes the positive and negative test suites, 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 January 2027. 10 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5.5 | 92.18 |
| 2 | Claude Opus 4.7 | 90.17 |
| 3 | Gemini 3.1 Pro (Preview) | 88.79 |
| 4 | Claude Sonnet 4.6 | 88.37 |
| 5 | DeepSeek V4 Pro | 87 |
| 6 | DeepSeek V4 Flash | 79.49 |
| 7 | GPT-5.4 Mini | 78.75 |
| 8 | Qwen 3.6 27B | 77.91 |
| 9 | Gemini 3 Flash (Preview) | 49.15 |
| 10 | GLM-4.7 Flash | 41.12 |
Interactive version: theaggregate.ai/benchmark?slug=vericontest-codegen-natural-language · How It Works · Data refreshed daily, snapshot 2026-10-07.