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 | Claude Opus 5 (High) | Anthropic | +13.77% | +15.59% | 21.87% | 15.56% | 15% | 0.82% | 21,433 |
| 2 | Claude Opus 5 (Max) | Anthropic | +11.61% | +15.64% | 18.91% | 7.19% | 15.47% | 0.86% | 17,073 |
| 3 | Claude Fable 5 (High) | Anthropic | +10.58% | +7.74% | 18.9% | 13.37% | 11.98% | 0.88% | 34,905 |
| 4 | GPT 5.6 Sol (xHigh) | OpenAI | +9.76% | +7.89% | 22.82% | 8.38% | 8.82% | 0.89% | 28,524 |
| 5 | Claude Opus 4.8 (High) | Anthropic | +9.51% | +5.86% | 20.7% | 11.46% | 9.83% | -0.28% | 36,887 |
| 6 | Kimi K3 (Max) | Moonshot | +8.74% | +16.89% | 16.2% | 1.85% | 7.86% | 0.89% | 94,549 |
| 7 | GPT 5.5 (xHigh) | OpenAI | +7.89% | +2.61% | 13.34% | 8.74% | 13.88% | 0.88% | 50,040 |
| 8 | Claude Sonnet 5 (High) | Anthropic | +7.48% | +0.98% | 12.24% | 12.4% | 11.05% | 0.74% | 27,467 |
| 9 | Claude Opus 4.7 (High) | Anthropic | +6.64% | +4.09% | 10.61% | 6.39% | 11.29% | 0.8% | 36,570 |
| 10 | Claude Opus 4.7 | Anthropic | +6.33% | +3.65% | 10.63% | 7.86% | 8.66% | 0.83% | 37,121 |
| 11 | GPT 5.5 (High) | OpenAI | +6.10% | +1.68% | 10.65% | 5.87% | 11.4% | 0.89% | 73,880 |
| 12 | GLM 5.2 (Max) | Zai | +6.08% | +8.49% | 10.59% | 5.45% | 4.99% | 0.89% | 65,079 |
| 13 | Grok 4.5 | xAI | +6.06% | +6.01% | 5.63% | 7.2% | 10.57% | 0.89% | 34,008 |
| 14 | Qwen3.8 Max | Alibaba | +6.01% | +10.76% | 6.8% | 5.4% | 7.18% | -0.11% | 18,386 |
| 15 | DeepSeek V4 Pro (High) (0813) | DeepSeek | +5.90% | +12.90% | 5.39% | 1.23% | 9.1% | 0.89% | 23,928 |
| 16 | Grok 4.6 (xHigh) | xAI | +5.76% | +11.89% | 2.52% | 3.8% | 9.7% | 0.89% | 15,090 |
| 17 | Anthropic | +5.22% | +3.13% | 7.52% | 5.24% | 9.31% | 0.88% | 36,365 | |
| 18 | GPT 5.5 | OpenAI | +4.84% | +2.11% | 6.32% | 4.76% | 10.13% | 0.89% | 77,762 |
| 19 | GLM 5.3 Flash | Zai | +4.41% | +15.21% | 5.21% | 1.89% | -1.15% | 0.89% | 9,970 |
| 20 | GLM 5.3 (Max) | Zai | +3.82% | +12.60% | 4.23% | 0.73% | 0.65% | 0.89% | 41,022 |
| 21 | GPT 5.4 (High) | OpenAI | +3.25% | +2.52% | -0.11% | 4.37% | 8.57% | 0.89% | 76,973 |
| 22 | Deepseek V4 Flash (High) (20260731) | DeepSeek | +2.97% | +7.45% | 2.61% | 0.7% | 3.24% | 0.87% | 51,219 |
| 23 | GPT 5.6 Terra (xHigh) | OpenAI | +2.92% | -3.10% | -1.02% | 9.42% | 8.4% | 0.89% | 16,778 |
| 24 | Qwen3.8 Flash Next | Alibaba | +2.44% | +12.34% | -1.59% | -1.06% | 2.92% | -0.41% | 8,777 |
| 25 | Claude Opus 4.8 | Anthropic | +2.33% | +6.69% | 11.68% | 9.92% | 10.51% | -27.15% | 33,690 |
| 26 | GPT 5.6 Luna (xHigh) | OpenAI | +2.13% | -0.74% | -0.13% | 2.42% | 8.22% | 0.89% | 16,383 |
| 27 | Qwen 3.8 27B | Alibaba | +1.50% | +7.24% | 1.39% | 0.2% | -1.49% | 0.16% | 16,060 |
| 28 | Kimi K2.7 Code | Moonshot | +1.38% | +2.15% | 1.19% | 5.26% | -2.56% | 0.89% | 11,142 |
| 29 | Muse Spark 1.2 (xHigh) | Meta | +0.97% | +6.38% | -6.4% | -5.49% | 9.47% | 0.88% | 18,416 |
| 30 | Gemini 3.7 Flash (High) | +0.96% | +9.83% | -1.6% | -5.19% | 0.93% | 0.82% | 24,413 | |
| 31 | Anthropic | +0.95% | -4.48% | -0.86% | -1.31% | 10.61% | 0.8% | 38,385 | |
| 32 | Kimi K2.6 | Moonshot | +0.29% | -4.03% | 4.87% | 8.23% | -8.49% | 0.89% | 11,279 |
| 33 | DeepSeek V4 Pro | DeepSeek | +0.10% | -1.37% | -2.16% | -0.83% | 4.72% | 0.12% | 31,185 |
| 34 | Muse Spark 1.1 | Meta | -0.73% | +5.55% | -8.22% | -5.7% | 3.83% | 0.87% | 87,115 |
| 35 | Qwen3.7 Max | Alibaba | -1.36% | -1.84% | -6.33% | -2.49% | 3.63% | 0.24% | 35,069 |
| 36 | GLM 5.1 | Zai | -1.41% | +0.60% | -1.48% | -0.6% | -4.37% | -1.2% | 71,952 |
| 37 | Hy3 | Tencent | -1.64% | -3.72% | -2.39% | -0.44% | -0.22% | -1.42% | 23,436 |
| 38 | Qwen3.7 Plus | Alibaba | -2.70% | -4.22% | -10.89% | -3.07% | 4.8% | -0.12% | 18,748 |
| 39 | Gemini 3.5 Flash (High) | -2.85% | -0.95% | -1.06% | -7.22% | -4.96% | -0.03% | 94,447 | |
| 40 | Gemini 3.1 Pro Preview | -3.13% | -1.62% | 1.96% | -1.9% | -14.72% | 0.64% | 83,248 | |
| 41 | Gemini 3.6 Flash (High) | -3.52% | -1.82% | -6.82% | -4.17% | -5.63% | 0.85% | 16,850 | |
| 42 | Mimo V2.5 Pro | Xiaomi | -3.56% | -5.35% | -8.8% | -2.68% | -0.85% | -0.13% | 36,154 |
| 43 | Minimax M3 | MiniMax | -3.61% | -8% | -9.85% | -5.8% | 5.12% | 0.45% | 35,470 |
| 44 | Gemini 3.5 Flash (Medium) | -4.07% | -8.58% | -5.51% | -2.34% | -4.22% | 0.32% | 13,762 | |
| 45 | Inkling Small | Thinky | -5.30% | -17.38% | -17.09% | -3.94% | 11.73% | 0.18% | 10,179 |
| 46 | Mistral Medium 3.5 | Mistral | -6.12% | -12.89% | -12.28% | 0.03% | -3.13% | -2.34% | 7,677 |
| 47 | Inkling | Thinky | -7.02% | -13.91% | -17.94% | -11% | 7.41% | 0.37% | 39,537 |
| 48 | Solar Pro 4 | Upstage | -9.19% | -8.17% | -13.47% | -1.06% | -22.5% | -0.73% | 5,770 |
| 49 | Minimax M2.7 | MiniMax | -10.26% | -12.28% | -16.93% | -4.64% | -18.2% | 0.75% | 35,721 |
| 50 | Grok 4.3 (High) | xAI | -10.76% | -12.37% | -13.31% | -11.38% | -17.43% | 0.69% | 62,791 |
| 51 | Gemini 3.5 Flash Lite | -10.81% | -14.47% | -14.07% | -9.71% | -14.98% | -0.83% | 22,348 | |
| 52 | -11.24% | -8.24% | -10.3% | -9.8% | -26.7% | -1.16% | 83,413 | ||
| 53 | Grok Build 0.1 | xAI | -11.31% | -6.83% | -11.52% | -5.99% | -32.75% | 0.53% | 74,459 |
| 54 | Nemotron 3 Ultra | Nvidia | -13.57% | -15.51% | -14.74% | -8.92% | -28.95% | 0.27% | 12,264 |
| 55 | Grok 4.3 | xAI | -18.81% | -12.42% | -15.13% | -10.51% | -56.78% | 0.77% | 82,745 |
| 56 | Gemma 4 31B | -22.33% | -2.54% | -4.06% | -10.35% | -65.28% | -29.43% | 56,661 |
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.