MERA - ruDetox: leaderboard
MERA is the Russian open benchmark run by the AI Alliance Russia; this is its ruDetox task: detoxification: rewrite a toxic Russian sentence keeping its content. Each MERA task keeps its own native metric, so the boards are not on one shared scale.
Metric: Joint Score (%). Source: mera.a-ai.ru. Saturation forecast: Around December 2027. 631 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemma 4 31B (IT) | 40.72 | #176 |
| 2 | Claude Opus 4.6 | 39.25 | #63 |
| 3 | Claude Opus 4.5 | 39.25 | #83 |
| 4 | GPT-5.4 | 38.72 | #79 |
| 5 | Kimi K2.5 | 38.19 | #144 |
| 6 | Gemma 4 26B A4B (IT) | 38.19 | #297 |
| 7 | Llama 3.1 405B Instruct | 38.13 | #457 |
| 8 | GPT-4o | 37.88 | #343 |
| 9 | DeepSeek V3 (0324) | 37.74 | #340 |
| 10 | Claude 3.7 Sonnet | 37.41 | #249 |
| 11 | Qwen 2.5 72B Instruct | 37.21 | #454 |
| 12 | GPT-5.2 | 36.99 | #110 |
| 13 | DeepSeek V3 | 36.79 | #321 |
| 14 | GLM-5.1 | 36.45 | #97 |
| 15 | Mistral Small 3.1 | 35.55 | #610 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mera-rudetox · How It Works · Data refreshed daily, snapshot 2026-10-20.