VeriContest - CodeGen (Natural Language): 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 problem statement, checked against all positive tests with a one-second timeout; pass@1 in percent, higher is better.

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 February 2027. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1GPT-5.5 (2026-04-23)92.18#39
2Claude Opus 4.790.17#86
3Gemini 3.1 Pro (Preview)88.79#102
4Claude Sonnet 4.688.37#179
5DeepSeek V4 Pro87#201
6DeepSeek V4 Flash79.49#273
7GPT-5.4 Mini (2026-03-17)78.75#296
8Qwen 3.6 27B77.91#358
9Gemini 3 Flash (Preview)49.15#156
10GLM-4.7 Flash41.12#1349

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=vericontest-codegen-natural-language · How It Works · Data refreshed daily, snapshot 2026-11-22.