MERA Code - ruCodeReviewer: leaderboard
MERA Code (Russian) ruCodeReviewer task: producing a code review in Russian and being scored against human reviews. Each task keeps its own metric — pass rates, CodeBLEU and chrF all appear across this suite — so read the metric label before comparing boards.
Metric: Top-1 Accuracy (%). Source: mera.a-ai.ru. Saturation forecast: Around September 2028. 66 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Opus 4.6 | 17.13 | #63 |
| 2 | GLM-5.1 | 16.11 | #97 |
| 3 | Qwen 3.6 27B | 15.24 | #170 |
| 4 | Kimi K2.5 | 14.51 | #144 |
| 5 | GPT-5.1 Codex Max | 12.34 | #100 |
| 6 | GPT-5.4 | 12.05 | #79 |
| 7 | Qwen 3.6 35B A3B | 11.76 | #226 |
| 8 | Qwen 3.5 27B | 10.6 | #187 |
| 9 | DeepSeek V4 Pro | 9.87 | #102 |
| 10 | GPT-4.1 | 9.58 | #248 |
| 11 | Qwen 3 32B | 9.58 | #443 |
| 12 | Qwen 3.5 122B A10B | 9.58 | #173 |
| 13 | Qwen 3 Next 80B A3B (Thinking) | 9.29 | #320 (Qwen 3 Next 80B A3B) |
| 14 | Qwen 3.5 35B A3B | 9.14 | #256 |
| 15 | DeepSeek V4 Flash | 9.14 | #137 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mera-code-rucodereviewer · How It Works · Data refreshed daily, snapshot 2026-10-20.