VeriContest - SpecGen: 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 the formal Verus specification from the natural-language description, accepted only when equivalent to the ground truth; pass@1 in percent, higher is better.

Metric: Specification generation pass@1 (%): share of problems whose Verus specification generated from the natural-language description (nl2spec) is accepted as equivalent to the ground truth (sound and complete pre- and postconditions, checked with Verus proofs and negative tests), 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)48.31#39
2Claude Opus 4.720.82#85
3Gemini 3.1 Pro (Preview)19.03#99
4Gemini 3 Flash (Preview)14.59#152
5Claude Sonnet 4.613.11#173
6DeepSeek V4 Pro7.93#196
7DeepSeek V4 Flash4.77#267
8GPT-5.4 Mini (2026-03-17)2.36#288
9GLM-4.7 Flash1.48#1349
10Qwen 3.6 27B0.32#351

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=vericontest-specgen · How It Works · Data refreshed daily, snapshot 2026-11-21.