AnyModelAI Arena — рейтинг моделей
Живой рейтинг ИИ-моделей по тому, насколько хорошо они справляются с реальной агентной работой — успешность выполнения, управляемость, восстановление после сбоев и надёжность работы с инструментами. Сравните GPT, Claude, Gemini, DeepSeek, Grok и другие, а затем запустите любую из них через один API.
Общий рейтинг Agent Arena — меньше — лучше.
| # | Модель | Лаборатория | Общий | Успешность | Похвала | Управляемость | Восстановление | Галлюц. | Сессии |
|---|---|---|---|---|---|---|---|---|---|
| 1 | GPT 6.1 Sol (Max) | OpenAI | +14.78% | +15.69% | 34.49% | 14.73% | 4.53% | 0.65% | 12,705 |
| 2 | Claude Fable 5.1 (Max) | Anthropic | +14.12% | +15.71% | 27.14% | 14.7% | 11.6% | 0.57% | 16,282 |
| 3 | Gemini 4 Argon (High) | +13.73% | +14.58% | 31.29% | 16.78% | -21.41% | 0.02% | 28,097 | |
| 4 | GPT 6 Astra (Max) | OpenAI | +13.69% | +12.55% | 28.55% | 14.56% | 5.53% | 0.59% | 12,892 |
| 5 | Claude Sonnet 5.5 (Max) | Anthropic | +12.55% | +19.98% | 17.02% | 12.35% | 15.04% | 0.46% | 7,446 |
| 6 | GPT 6 Sol (Max) | OpenAI | +11.20% | +9.21% | 22.89% | 12.41% | 7.03% | 0.39% | 7,075 |
| 7 | Claude Opus 5 (Max) | Anthropic | +9.20% | +12.65% | 14.95% | 7.26% | 13.57% | 0.6% | 23,614 |
| 8 | Claude Fable 5 (High) | Anthropic | +8.39% | +5.05% | 15.07% | 10.59% | 9.6% | 0.62% | 42,474 |
| 9 | GPT 5.6 Sol (xHigh) | OpenAI | +7.42% | +3.75% | 14.75% | 8.42% | 0.73% | 0.51% | 36,373 |
| 10 | Claude Opus 4.8 (High) | Anthropic | +7.14% | +3.96% | 12.93% | 7.02% | 7.22% | 0.35% | 40,277 |
| 11 | Claude Sonnet 5 (High) | Anthropic | +4.69% | +1.58% | 4.74% | 6.58% | 7.38% | 0.41% | 33,161 |
| 12 | Gemini 3.8 Flash (High) | +3.97% | +5.08% | 4.55% | 2.09% | -0.69% | 0.52% | 37,462 | |
| 13 | MiMo V2.6 Pro | Xiaomi | +3.81% | +8.61% | 4.54% | 2% | 2.34% | -0.13% | 23,636 |
| 14 | Muse Spark 1.3 (Max) | Meta | +3.63% | +4.87% | 4.1% | 5.6% | 6.08% | 0.55% | 73,452 |
| 15 | Grok 4.7 (xHigh) | xAI | +3.39% | +6.12% | 3.48% | 0.62% | 7.2% | 0.56% | 14,843 |
| 16 | GPT 5.5 (xHigh) | OpenAI | +3.31% | -1.98% | 6.52% | 3.2% | 9.55% | 0.5% | 54,310 |
| 17 | GPT 6 Luna (Max) | OpenAI | +3.26% | +1.30% | 1.34% | 6.17% | 5.21% | 0.42% | 155,783 |
| 18 | Deepseek V4.1 Flash (Max) | DeepSeek | +2.95% | +7.48% | 3.58% | 1.94% | 7.86% | 0.46% | 228,658 |
| 19 | Hy4 preview | Tencent | +2.90% | +6.04% | 5.44% | 2.43% | 6.51% | 0.16% | 82,690 |
| 20 | MiMo V2.6 Flash | Xiaomi | +2.53% | +7.05% | 3.55% | 4.8% | -1.06% | -3.18% | 18,936 |
| 21 | GLM 5.2 (Max) | Zai | +2.35% | +3.55% | 6.76% | 1.23% | 0.67% | 0.62% | 86,843 |
| 22 | GLM 5.3 (Max) | Zai | +2% | +3.80% | 6.09% | 0.41% | -0.05% | 0.59% | 86,255 |
| 23 | Kimi K3 (Max) | Moonshot | +1.95% | +6.32% | 2.21% | 0.53% | 4.38% | 0.65% | 134,561 |
| 24 | Qwen3.8 Max | Alibaba | +1.92% | +4.69% | 2.09% | 1.91% | 2.8% | -0.77% | 67,520 |
| 25 | GPT 5.4 (High) | OpenAI | +1.64% | -4.28% | -0.91% | 7.1% | 1.21% | 0.34% | 81,779 |
| 26 | Grok 4.6 (xHigh) | xAI | +1.16% | -0.65% | -0.32% | 1.05% | 5.68% | 0.52% | 28,687 |
| 27 | GPT 5.6 Terra (xHigh) | OpenAI | +0.43% | -4.13% | 1.96% | 2.57% | 0.88% | 0.4% | 24,138 |
| 28 | DeepSeek V4 Pro (High) (0813) | DeepSeek | +0.36% | -1.45% | 0.72% | 0.67% | 4.4% | 0.48% | 7,071 |
| 29 | Grok 4.5 | xAI | +0.07% | +0.42% | -2.72% | 2.63% | 5.14% | 0.59% | 41,828 |
| 30 | GPT 5.5 | OpenAI | -0.07% | -6.08% | 2.82% | -0.98% | 4.12% | 0.43% | 82,556 |
| 31 | Gemini 3.7 Flash (High) | -0.41% | -0.76% | -2.09% | -3.21% | -5.44% | 0.4% | 54,750 | |
| 32 | GLM 5.3 Flash | Zai | -1.09% | +3.87% | -1.37% | -2.98% | -5.7% | 0.58% | 86,532 |
| 33 | Step 5 Preview | Stepfun | -1.33% | -0.10% | -2.21% | 1.01% | 0.23% | 0.64% | 13,472 |
| 34 | GPT 5.6 Luna (xHigh) | OpenAI | -1.80% | -5.99% | -4.75% | -1.47% | 0.51% | 0.43% | 43,217 |
| 35 | Qwen3.8 Flash Next | Alibaba | -2.56% | +2.65% | -3.2% | -0.94% | -2.57% | -0.73% | 97,147 |
| 36 | Qwen 3.8 27B | Alibaba | -3.21% | -0.56% | -5.06% | -3.6% | -5.9% | -0.22% | 60,184 |
| 37 | Gemini 3.1 Pro Preview | -3.53% | -10.09% | -1.06% | -6.17% | -23.27% | -0.2% | 87,156 | |
| 38 | Gemini 3.6 Flash (High) | -4.72% | -6.14% | -7.83% | -8.16% | -18.23% | 0.62% | 22,826 | |
| 39 | Qwen3.7 Max | Alibaba | -5.80% | -8.65% | -10.28% | -10.33% | -1.56% | -0.07% | 41,869 |
| 40 | Hy3 | Tencent | -6.13% | -10.49% | -6.05% | 0.25% | -7.21% | -4.16% | 31,121 |
| 41 | Muse Spark 1.1 | Meta | -6.25% | -4.74% | -10.97% | -11.13% | -2.26% | 0.11% | 145,611 |
| 42 | Qwen3.7 Plus | Alibaba | -6.35% | -8.08% | -9.61% | -9.21% | -6.31% | -1.33% | 23,173 |
| 43 | Muse Spark 1.2 (xHigh) | Meta | -6.52% | -7.85% | -11.34% | -8.35% | 5.79% | 0.19% | 51,136 |
| 44 | Mimo V2.5 Pro | Xiaomi | -7.67% | -8.98% | -13.38% | -11.55% | -7.51% | -1.07% | 48,270 |
| 45 | Minimax M3 | MiniMax | -7.86% | -13.56% | -13.66% | -9.16% | 0.58% | -0.12% | 42,680 |
| 46 | Mistral Medium 3.5 | Mistral | -11.33% | -15.40% | -16.84% | -11.33% | -20.45% | -3.24% | 12,602 |
| 47 | Inkling | Thinky | -13.61% | -22.79% | -22.61% | -16.69% | 0.08% | -0.52% | 43,418 |
| 48 | Inkling Small | Thinky | -14.72% | -25.08% | -22.69% | -19.63% | 5.43% | -1.34% | 13,952 |
| 49 | Solar Pro 4 | Upstage | -20.18% | -26.63% | -21.8% | -14.53% | -25.56% | -2.1% | 10,158 |
Источник данных: Agent Arena (arena.ai) · lmarena-ai/leaderboard-dataset · CC BY 4.0 · обновляется ежедневно
Как работает рейтинг
Каждая модель оценивается на реальных агентных сессиях: выполняет ли она задачу, насколько хорошо принимает правки (управляемость), насколько надёжно восстанавливается после неудачных команд и как часто выдумывает несуществующие инструменты (чем меньше, тем лучше). Переключите ракурс выше, чтобы отсортировать по любому из этих показателей.
Сравнение AI-моделей (арена ИИ)
Рейтинг сравнивает нейросети — GPT, Claude, Gemini, DeepSeek, Grok и другие — по реальным агентным задачам: успешность выполнения, управляемость, восстановление после ошибок и надёжность работы с инструментами. Любую модель из таблицы можно запустить через один API AnyModel.
Запускайте топовые модели через один API
Без подписок — оплата за токены, бесплатный старт. Переключайтесь между любыми моделями из таблицы, меняя один id.