CVerifBench: leaderboard
We introduce an evaluation framework of 500 C verification tasks across five property types (memory safety, overflow, termination, reachability, data races) built on SV-COMP 2025, and evaluate 14 models across six families. We find that high overall accuracy masks a critical weakness: while most models reliably confirm properties hold, violation detection varies widely and degrades sharply with pr
Metric: Total (self-reported). Source: benchmarklist.com. Saturation forecast: Around June 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.7 | 98.3 | #49 |
| 2 | Kimi K2.5 | 97.7 | #144 |
| 3 | DeepSeek V3.2 | 95.8 | #205 |
| 4 | Qwen 3 32B | 95.4 | #443 |
| 5 | Claude Sonnet 4.5 | 94.9 | #142 |
| 6 | Ministral 3 8B | 93.8 | #687 |
| 7 | DeepSeek R1 | 93.7 | #251 |
| 8 | Qwen 3 Coder 480B A35B | 93.7 | #279 |
| 9 | GPT-OSS-120B | 91.9 | #341 |
| 10 | Qwen 3 8B | 90.7 | #674 |
| 11 | GPT-OSS-20B | 84.8 | #518 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=cverifbench · How It Works · Data refreshed daily, snapshot 2026-10-20.