LiveFMBench - 2025: leaderboard
Metric: Pass@1 (%) after the faithfulness check on LiveFMBench-2025 (360 C programs from SV-COMP 2025), direct prompting (instructions plus the C program), 32 samples at temperature 0.7 and top-p 0.95, pass@1 estimated from the 32 samples with the unbiased estimator; a sample counts only when Frama-C/WP (Alt-Ergo and Z3) proves the program and its code and assertions are unchanged (faithfulness check); higher is better. Source: arxiv.org. Saturation forecast: Around October 2027. 15 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | DeepSeek R1 0528 | 15.73 |
| 2 | DeepSeek V3.1 Terminus (Thinking) | 13.64 |
| 3 | GLM-4.6 | 12.3 |
| 4 | Qwen 3 32B | 7.44 |
| 5 | Qwen 3 235B A22B 2507 (Thinking) | 7.38 |
| 6 | GLM-4.6 (Non-reasoning) | 7.37 |
| 7 | Kimi K2 0905 | 6.32 |
| 8 | DeepSeek V3.2 (Thinking) | 4.65 |
| 9 | Qwen 3 Coder 480B A35B Instruct | 4.55 |
| 10 | DeepSeek V3.1 Terminus | 2.87 |
| 11 | DeepSeek V3.2 (Non-reasoning) | 2.13 |
| 12 | Qwen 3 235B A22B 2507 Instruct | 1.89 |
| 13 | DeepSeek V3 (0324) | 1.28 |
| 14 | Qwen 3 32B (Non-reasoning) | 0.29 |
| 15 | Llama 3.3 70B Instruct | 0.25 |
Interactive version: theaggregate.ai/benchmark?slug=livefmbench-2025 · How It Works · Data refreshed daily, snapshot 2026-10-07.