VeriContest - CodeGen (Natural Language and Specification): leaderboard
VeriContest (2026) is a verifiable code generation benchmark of 946 LeetCode and Codeforces problems in Rust with Verus, each with expert-validated specifications, reference code and Verus-checked proofs. This task asks for executable Rust code from the natural-language statement together with the formal specification, checked against the test suites; pass@1 in percent, higher is better.
Metric: Code generation pass@1 (%) from the natural-language description plus the ground-truth Verus specification (nl_spec2code): share of problems whose generated Rust code passes the 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 February 2027. 10 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 77.7 | #99 |
| 2 | Claude Sonnet 4.6 | 75.59 | #173 |
| 3 | GPT-5.5 (2026-04-23) | 74.52 | #39 |
| 4 | Claude Opus 4.7 | 72.3 | #85 |
| 5 | DeepSeek V4 Pro | 67.44 | #196 |
| 6 | DeepSeek V4 Flash | 60.19 | #267 |
| 7 | GPT-5.4 Mini (2026-03-17) | 58.14 | #288 |
| 8 | Qwen 3.6 27B | 48.94 | #351 |
| 9 | Gemini 3 Flash (Preview) | 34.99 | #152 |
| 10 | GLM-4.7 Flash | 21.46 | #1349 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=vericontest-codegen-natural-language-and-specification · How It Works · Data refreshed daily, snapshot 2026-11-21.