PAIR-Bench - Targeted Repair: leaderboard

Metric: Targeted repair success (%; share of the hidden tests in the hinted failure scenario that pass after the repair turn, averaged over turns and initially failing instances; higher is better). Source: arxiv.org. Saturation forecast: Around December 2026. 8 models tracked.

Top models

#ModelScore
1DeepSeek V3.271.59
2Gemini 2.5 Flash Lite59
3GPT-4o Mini56.14
4Qwen 3 Coder 30B A3B Instruct55.89
5Llama 3.3 70B Instruct53.5
6Ministral 3 14B49.71
7Mistral Small 3.249.31
8Gemma 3 27B42.56

Interactive version: theaggregate.ai/benchmark?slug=pair-bench-targeted-repair · How It Works · Data refreshed daily, snapshot 2026-09-29.