MERA Code - RealCodeJava: leaderboard
Metric: pass@1 (%). Source: mera.a-ai.ru. 66 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | GPT-4.1 | 41.61 |
| 2 | Claude Opus 4.6 | 40.94 |
| 3 | Gemini 2.5 Flash | 38.59 |
| 4 | Qwen 2.5 Coder 32B Instruct | 38.59 |
| 5 | DeepSeek V4 Flash | 37.25 |
| 6 | Mixtral 8x22B Instruct (v0.1) | 36.58 |
| 7 | GPT-5.4 | 36.24 |
| 8 | GLM-5.1 | 36.24 |
| 9 | Kimi K2.5 | 35.91 |
| 10 | GPT-5.1 Codex Max | 35.91 |
| 11 | Qwen 2.5 72B Instruct | 34.9 |
| 12 | DeepSeek V4 Pro | 34.56 |
| 13 | Kimi K2 0905 | 34.23 |
| 14 | Qwen 3 Coder 480B A35B Instruct | 34.23 |
| 15 | GPT-OSS-120B | 30.87 |
Interactive version: theaggregate.ai/benchmark?slug=mera-code-realcodejava · How It Works · Data refreshed daily, snapshot 2026-09-05.