MERA Code - RealCode: leaderboard
MERA Code (Russian) RealCode task: completing functions inside real repositories rather than isolated prompts. 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: pass@1 (%). Source: mera.a-ai.ru. Saturation forecast: Around January 2027. 66 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-5.4 | 54.24 | #79 |
| 2 | GPT-5.1 Codex Max | 48.5 | #100 |
| 3 | GLM-5.1 | 45.51 | #97 |
| 4 | Kimi K2.5 | 44.76 | #144 |
| 5 | Claude Opus 4.6 | 44.14 | #63 |
| 6 | Qwen 3.6 27B | 43.27 | #170 |
| 7 | Qwen 3.5 27B | 43.02 | #187 |
| 8 | GPT-4.1 | 41.77 | #248 |
| 9 | DeepSeek V4 Pro | 41.77 | #102 |
| 10 | Qwen 3 Coder 480B A35B Instruct | 41.15 | #315 |
| 11 | Qwen 3.6 35B A3B | 41.02 | #226 |
| 12 | Qwen 3.5 122B A10B | 40.27 | #173 |
| 13 | Gemini 2.5 Flash | 38.78 | #245 |
| 14 | Qwen 3.5 35B A3B | 38.65 | #256 |
| 15 | DeepSeek V4 Flash | 37.91 | #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-realcode · How It Works · Data refreshed daily, snapshot 2026-10-20.