Ru Arena Hard: leaderboard
Russian-language Arena Hard Auto: pairwise comparison of LLM responses to Russian prompts using GPT-4 as judge, reporting win rates.
Metric: Win Rate (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 137 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek V3 Chat | 96.3 | #447 |
| 2 | GPT-4o ChatGPT | 94.75 | |
| 3 | Yi Lightning | 93.46 | #423 |
| 4 | O1 Mini | 93.45 | #323 |
| 5 | Claude 3 Opus (20240229) | 91.3 | #412 |
| 6 | GPT-4 Preview (1106) | 90.89 | #418 |
| 7 | O1 Preview | 90.8 | #193 |
| 8 | Gemini 1.5 Pro (002) | 89.08 | #349 |
| 9 | Qwen 2.5 72B Instruct | 88.25 | #454 |
| 10 | Claude 3.5 Sonnet (20240620) | 88.17 | #299 |
| 11 | Gemma 2 27B (IT) | 86.87 | #645 |
| 12 | Phi-4 | 86.59 | #706 |
| 13 | Claude 3 Sonnet (20240229) | 84.27 | #677 |
| 14 | GPT-4o Mini | 83.9 | #599 |
| 15 | Llama 3.1 70B Instruct | 83.26 | #561 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ru-arena-hard · How It Works · Data refreshed daily, snapshot 2026-10-20.