CodeSpecBench-Func: leaderboard
Metric: Pass rate (%), the share of generated executable pre/postcondition specifications that are both correct (accept every valid test) and complete (reject every invalid test), on CodeSpecBench-Func (2,494 LeetCode problems; 50 valid and 50 invalid inputs plus valid and model-generated buggy outputs per problem, validated by the online judge); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-5 Mini | 47 |
| 2 | Gemini 2.5 Pro | 46.2 |
| 3 | GPT-OSS-120B | 42.5 |
| 4 | Claude Sonnet 4.5 (Thinking) | 40.5 |
| 5 | GPT-5 | 40.2 |
| 6 | GPT-OSS-20B | 37.7 |
| 7 | Gemini 2.5 Flash | 36.5 |
| 8 | QwQ-32B | 28.7 |
| 9 | Qwen 3 14B | 26.4 |
| 10 | Qwen 3 32B | 25.8 |
| 11 | DeepSeek V3.2 (Non-reasoning) | 24.3 |
| 12 | Qwen 3 32B (Non-reasoning) | 17.6 |
| 13 | Qwen 3 8B | 16.8 |
| 14 | Qwen 3 4B | 15.5 |
| 15 | Qwen 3 14B (Non-reasoning) | 13.5 |
Interactive version: theaggregate.ai/benchmark?slug=codespecbench-func · How It Works · Data refreshed daily, snapshot 2026-10-07.