MCR-Bench - C# Defect Identification F1: leaderboard
MCR-Bench (arXiv 2608.27442v1, Table 5): multi-round code review, where a model follows pull-request discussions and code revisions across review rounds and must identify the defects raised. Defect identification F1 against annotated defects on the C# subset, percent; plain prompting without a review scaffold. Higher is better.
Metric: F1 (%; matched defects). Source: arxiv.org. Saturation forecast: Around January 2027. 7 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Haiku 4.5 | 54.8 | #280 |
| 2 | Gemini 3 Flash | 53.8 | #99 |
| 3 | GPT-5.2 | 53.3 | #110 |
| 4 | DeepSeek V3.2 | 49.1 | #205 |
| 5 | GLM-4.7 | 46.7 | #192 |
| 6 | Kimi K2 | 38.9 | #243 |
| 7 | Qwen 3 Max | 15.3 | #208 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mcr-bench-c-sharp-defect-identification-f1 · How It Works · Data refreshed daily, snapshot 2026-10-20.