| Модель | Контекст | Ввод | Вывод | Пропускная способность | Задержка | Провайдеры | Агенты | Кодинг | Рассуждения | Знания | Математика | Мультиязычность | Мультимодальность | Следование инструкциям |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GPT-4 | 8.2K | — | — | 35 t/s | 0.74с | +68 | — | — | — | — | — | — | — | — |
| GPT-3.5 Turbo Instruct | 4.1K | — | — | — | — | +34 | — | — | — | — | — | — | — | — |
| GPT-3.5 Turbo | 16.4K | — | — | 115 t/s | 2.21с | +86 | — | — | 38 | — | 44.1 | — | — | — |
| Claude 3 Haiku | 200K | — | — | — | — | +14 | 21.1 | 13.4 | 20.8 | 28.1 | 20.2 | — | 30.8 | 36.1 |
| GPT-4 Turbo | 128K | — | — | 50 t/s | 0.97с | +64 | — | 28.6 | 36.3 | 5.6 | 44.3 | — | — | — |
| WizardLM-2 8x22B | 65.5K | — | — | — | — | +4 | — | — | — | — | — | — | — | — |
| GPT-4o Mini | 128K | — | — | 84 t/s | 4.10с | +108 | 50.5 | 20.1 | 37.1 | 17.8 | 45.3 | — | 41.5 | 31 |
| Mistral Nemo | 131.1K | — | — | 41 t/s | 2.15с | +8 | — | — | — | — | — | — | — | — |
| Hermes 3 Llama 3.1 | 131.1K | — | — | — | — | +17 | — | — | — | — | — | — | — | — |
| Claude 3.5 Haiku | 200K | — | — | 42 t/s | 2.46с | +26 | — | 29.4 | 35.9 | — | 37.7 | — | — | — |
| GPT-4o | 128K | — | — | 82 t/s | 3.44с | +119 | 25.1 | 26.5 | 44.1 | 25.3 | 45.5 | — | 100 | 34.3 |
| O1 | 200K | — | — | — | — | +64 | 62.6 | 38.4 | 45.2 | 50.3 | 84.7 | — | — | 81.3 |
| Phi 4 | 16.4K | — | — | 43 t/s | 1.32с | +13 | — | 19.7 | 44.3 | 32 | 47.7 | — | — | 23.5 |
| MiniMax 01 | 1.0M | — | — | — | — | +4 | — | — | — | — | — | — | — | — |
| DeepSeek R1 Distill Qwen | 128K | — | — | 50 t/s | 19.88с | +47 | — | 34.1 | 37.7 | 26 | 81.4 | — | — | 22.9 |
| Aion RP Llama 3.1 | 32.8K | — | — | — | — | +1 | — | — | — | — | — | — | — | — |
| O3 Mini | 200K | — | — | 64 t/s | 6.90с | +74 | 28.7 | 41.5 | 54.2 | 47.3 | 87.2 | — | — | 93.9 |
| GPT-4o Search | 128K | — | — | 61 t/s | 3.47с | +36 | — | — | — | — | — | — | — | — |
| GPT-4o Mini Search | 128K | — | — | — | — | +33 | — | — | — | — | — | — | — | — |
| O1 Pro | 200K | — | — | — | — | +48 | — | — | 79 | 18.9 | — | — | — | — |
| Llama 4 Scout | 10M | — | — | 75 t/s | 1.62с | +32 | 31.9 | 14.7 | 41 | 33.2 | 56.4 | — | 76.5 | 39.5 |
| GPT-4.1 Nano | 1.0M | — | — | 140 t/s | 1.78с | +91 | 16.8 | 19.9 | 34.5 | 39.7 | 54.2 | — | 70.1 | 57.6 |
| GPT-4.1 Mini | 1.0M | — | — | 93 t/s | 2.69с | +100 | 51.5 | 30.1 | 47.8 | 45.5 | 67.8 | — | 79.4 | 63.4 |
| GPT-4.1 | 1.0M | — | — | 85 t/s | 1.92с | +98 | 36.4 | 38 | 53.2 | 47.4 | 67.5 | — | 80.6 | 65.2 |
| O3 | 200K | — | — | 134 t/s | 2.82с | +74 | 80.7 | 50 | 51.7 | 51.7 | 94.8 | — | 85.1 | 71.4 |
| O4 Mini | 200K | — | — | 94 t/s | 8.08с | +77 | — | 66.2 | 59.7 | — | 96.4 | — | — | — |
| Claude Sonnet 4 | 1M | — | — | 34 t/s | 4.93с | +108 | 46 | 43.9 | 40.3 | 32.2 | 67 | — | 81.2 | 45.4 |
| Claude Opus 4 | 200K | — | — | 25 t/s | 12.05с | +78 | — | 47.6 | 54 | — | 75.2 | — | — | — |
| DeepSeek R1 | 163.8K | — | — | 52 t/s | 10.36с | +138 | 36.5 | 43.4 | 47.4 | 46.3 | 93.8 | — | — | 39.6 |
| O3 Pro | 200K | — | — | — | — | +52 | — | — | 84.5 | 58.5 | — | — | — | — |
| Gemini 2.5 Pro | 1.0M | — | — | 88 t/s | 16.94с | +153 | 42.2 | 41.4 | 52.1 | 51.5 | 92.7 | — | 87.5 | 48.7 |
| Gemini 2.5 Flash | 1.0M | — | — | 154 t/s | 8.62с | +147 | 14.9 | 37.5 | 44.1 | 41.5 | 90.2 | — | 82.8 | 39 |
| MiniMax M1 | 1M | — | — | — | — | +23 | — | 54.3 | 53.2 | — | 91.3 | — | — | — |
| Qwen3 | 262.1K | — | — | 100 t/s | 12.55с | +116 | — | 41 | 50.1 | 62.6 | 85.1 | 79.4 | — | — |
| Qwen3 Coder | 1.0M | — | — | 243 t/s | 5.28с | +53 | — | — | — | — | — | — | — | — |
| GLM-4.5 Air | 131.1K | — | — | 60 t/s | 10.99с | +84 | 46.5 | 37.5 | 51.3 | 40.9 | 81.9 | — | 100 | 37.6 |
| GLM-4.5 | 131.1K | — | — | 39 t/s | 7.22с | +69 | — | 54.3 | 58 | — | 92.6 | — | 100 | — |
| Claude Opus 4.1 | 200K | — | — | 22 t/s | 3.56с | +89 | — | 74.5 | — | 28.2 | — | — | 100 | — |
| GPT-OSS | 131.1K | — | — | 393 t/s | 2.28с | +124 | — | — | — | — | — | — | — | — |
| GPT-5 Nano | 400K | — | — | 123 t/s | 10.60с | +98 | — | 57.8 | 51.3 | — | — | — | — | — |
| GPT-5 Mini | 400K | — | — | 90 t/s | 7.03с | +99 | — | 45.7 | 62.1 | — | — | — | — | — |
| GPT-5 | 400K | — | — | 88 t/s | 6.23с | +142 | 59.8 | 38 | 46.9 | 53.9 | 61.4 | — | 87.1 | 73.1 |
| GLM-4.5V | 65.5K | — | — | 73 t/s | 6.15с | +32 | — | 27 | 45.3 | — | — | — | — | — |
| Mistral Medium 3.1 | 131.1K | — | — | — | — | +1 | — | 37.2 | 43.8 | — | — | — | — | — |
| DeepSeek Chat V3.1 | 163.8K | — | — | 23 t/s | 2.02с | +5 | — | — | — | — | — | — | — | — |
| Kimi K2 | 131.1K | — | — | 29 t/s | 2.21с | +79 | 61.1 | 35.1 | 54.3 | 40.8 | 83.2 | — | 100 | 41.5 |
| DeepSeek V3.1 Terminus | 163.8K | — | — | 58 t/s | 1.50с | +81 | — | 42.5 | 55.7 | — | — | — | — | — |
| GPT-5 Codex | 400K | — | — | 154 t/s | 2.76с | +117 | — | 62.5 | 65.3 | — | — | — | — | — |
| Qwen3 Coder Plus | 1M | — | — | 45 t/s | 1.35с | +64 | — | — | — | — | — | — | — | — |
| Qwen3 Max | 262.1K | — | — | 38 t/s | 7.37с | +83 | 59 | 32.9 | 54.1 | 45.1 | — | — | 100 | 44.1 |
| Gemini 2.5 Flash Lite | 1.0M | — | — | 277 t/s | 1.41с | +106 | — | 39.3 | 48.3 | — | 83.6 | — | — | — |
| Claude Sonnet 4.5 | 1M | — | — | 41 t/s | 4.57с | +154 | 44.2 | 59.7 | 44.8 | — | 87 | — | 100 | — |
| GLM-4.6 | 202.8K | — | — | 53 t/s | 16.68с | +89 | 76.9 | 34.6 | 50.1 | 35.7 | — | — | — | 36.7 |
| GPT-5 Pro | 400K | — | — | — | — | +61 | — | — | — | — | — | — | — | — |
| O4 Mini Deep Research | 200K | — | — | — | — | +21 | — | — | — | — | — | — | — | — |
| O3 Deep Research | 200K | — | — | — | — | +23 | — | — | — | — | — | — | — | — |
| Claude Haiku 4.5 | 200K | — | — | 102 t/s | 2.77с | +172 | — | 52.9 | 49.6 | — | — | — | 100 | — |
| Phi 4 Mini Instruct | 131.1K | — | — | — | — | +24 | — | 11.7 | 27.9 | — | 36.3 | — | — | — |
| Granite 4.0 H Micro | 131K | — | — | — | — | +4 | — | — | — | — | — | — | — | — |
| MiniMax M2 | 204.8K | — | — | 48 t/s | 10.12с | +70 | — | 59.4 | 57.4 | — | — | — | — | — |