PAIR-Bench - Initial Fix: leaderboard
The zero-hint protocol of PAIR-Bench (2026): the percent of 440 wrong-answer Codeforces Python programs a model repairs to pass every hidden test in its first attempt, before any feedback. Higher is better.
Metric: Initial fix rate (%; share of the 440 wrong-answer programs a model repairs to pass every hidden test in its first attempt, before any hint; higher is better). Source: arxiv.org. Saturation forecast: Around February 2027. 8 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V3.2 | 65.68 | #406 |
| 2 | Gemini 2.5 Flash Lite | 53.18 | #958 |
| 3 | Qwen 3 Coder 30B A3B Instruct | 35.9 | #972 |
| 4 | GPT-4o Mini | 29.77 | #1623 |
| 5 | Llama 3.3 70B Instruct | 27.72 | #1357 |
| 6 | Ministral 3 14B | 22.95 | #1845 |
| 7 | Mistral Small 3.2 | 13.4 | #1591 |
| 8 | Gemma 3 27B | 10.22 | #1631 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=pair-bench-initial-fix · How It Works · Data refreshed daily, snapshot 2026-11-13.