Рейтинг текстовых моделей
Таблица по открытым замерам оценок людей: оценка, интервал и число голосов; переключайте категорию и дату снимка.
Динамика рейтинга
Загружаем историю замеров…
Как считается
Оценки и интервалы приведены как в открытых замерах оценок людей, без пересчёта; доля использования — по публичным данным API-агрегаторов (топ-K моделей и строка «прочие»); характеристики и цены — из каталогов и провайдеров с датой наблюдения. Интервал у оценки не равен диапазону мест: соседние модели с пересекающимися интервалами статистически неразличимы.
Полная методика: источники, частота обновления, что означает число
Смысл числа
- Метрика
- рейтинг Брэдли — Терри (Elo-подобная шкала) с интервалом из источника
- Единица выборки
- голос в парном сравнении
- Окно
- снимок датасета от 13 сентября 2026
- Как читать
- интервал не равен диапазону мест: модели с пересекающимися интервалами статистически неразличимы
Таблица борда
Снимок от 13 сентября 2026, категория «сложные запросы (англ.)», 300 строк. Оценка и интервал — из источника; интервал не равен диапазону мест.
| Место | Модель | Разработчик | Оценка | Интервал | Голоса | Лицензия |
|---|---|---|---|---|---|---|
| 1 | Claude Opus 4.6 в режиме high | Anthropic | 1 531 | 1 526–1 537 | 20 829 | Proprietary |
| 2 | Claude Opus 4.6 | Anthropic | 1 527 | 1 522–1 532 | 22 818 | Proprietary |
| 3 | Claude Opus 5 в режиме max | Anthropic | 1 522 | 1 514–1 531 | 5 521 | Proprietary |
| 4 | Claude Fable 5.1 в режиме max | Anthropic | 1 519 | 1 503–1 536 | 1 416 | Proprietary |
| 5 | Claude Opus 5 в режиме high | Anthropic | 1 519 | 1 512–1 526 | 11 521 | Proprietary |
| 6 | Gemini 3.8 Flash в режиме high | 1 512 | 1 496–1 528 | 1 348 | Proprietary | |
| 7 | Claude Fable 5 | Anthropic | 1 510 | 1 503–1 518 | 8 708 | Proprietary |
| 8 | Claude Opus 4.7 в режиме high | Anthropic | 1 506 | 1 501–1 512 | 19 416 | Proprietary |
| 9 | GLM 5.3 в режиме max | Z.ai | 1 505 | 1 493–1 516 | 2 919 | MIT |
| 10 | Qwen3.8 Max (0902) | Alibaba Qwen | 1 504 | 1 494–1 513 | 4 428 | Proprietary |
| 11 | Claude Opus 4.7 | Anthropic | 1 501 | 1 495–1 507 | 19 616 | Proprietary |
| 12 | MiMo-V2.5-Pro | Xiaomi | 1 496 | 1 490–1 502 | 17 782 | MIT |
| 13 | Kimi K3 в режиме max | Moonshot AI | 1 495 | 1 486–1 503 | 5 277 | Kimi K3 license |
| 14 | Claude Sonnet 4.6 | Anthropic | 1 494 | 1 488–1 499 | 20 001 | Proprietary |
| 15 | Gemini 3.7 Flash в режиме high | 1 490 | 1 475–1 505 | 1 599 | Proprietary | |
| 16 | Muse Spark 1.3 в режиме max | Meta | 1 490 | 1 473–1 506 | 1 285 | Proprietary |
| 17 | Muse Spark 1.2 в режиме xhigh | Meta | 1 488 | 1 468–1 507 | 907 | Proprietary |
| 18 | ERNIE 5.1 | Baidu | 1 485 | 1 479–1 492 | 11 680 | Proprietary |
| 19 | Claude Opus 4.8 в режиме high | Anthropic | 1 485 | 1 479–1 491 | 15 765 | Proprietary |
| 20 | GLM 5.3 Flash | Z.ai | 1 485 | 1 473–1 497 | 2 537 | MIT |
| 21 | Gemini 3.5 Flash в режиме high | 1 483 | 1 477–1 490 | 11 205 | Proprietary | |
| 22 | Qwen3.5 Max | Alibaba Qwen | 1 483 | 1 475–1 491 | 6 406 | Proprietary |
| 23 | GPT-5.5 в режиме high | OpenAI | 1 482 | 1 477–1 488 | 19 923 | Proprietary |
| 24 | Gemini 3.6 Flash в режиме high | 1 482 | 1 474–1 490 | 7 671 | Proprietary | |
| 25 | Muse Spark 1.1 | Meta | 1 482 | 1 474–1 489 | 7 653 | Proprietary |
| 26 | Claude Opus 4.5 | Anthropic | 1 482 | 1 477–1 487 | 18 070 | Proprietary |
| 27 | GPT-5.5 · 5.5 | OpenAI | 1 482 | 1 476–1 487 | 20 643 | Proprietary |
| 28 | GPT-5.4 в режиме high | OpenAI | 1 481 | 1 475–1 486 | 18 525 | Proprietary |
| 29 | GLM 5.1 | Z.ai | 1 480 | 1 474–1 486 | 14 544 | MIT |
| 30 | Gemini 3.1 Pro Preview | 1 479 | 1 474–1 484 | 31 868 | Proprietary | |
| 31 | Gemini 3.5 Flash в режиме medium | 1 478 | 1 471–1 485 | 10 699 | Proprietary | |
| 32 | Claude Opus 4.5 в режиме high | Anthropic | 1 477 | 1 470–1 485 | 7 605 | Proprietary |
| 33 | Muse Spark | Meta | 1 477 | 1 467–1 486 | 4 353 | Proprietary |
| 34 | Gemini 3 Pro | 1 476 | 1 469–1 482 | 8 866 | Proprietary | |
| 35 | GLM 5.2 в режиме max | Z.ai | 1 475 | 1 468–1 482 | 10 513 | MIT |
| 36 | Claude Opus 4.8 | Anthropic | 1 475 | 1 469–1 481 | 16 344 | Proprietary |
| 37 | Qwen3.7 Max | Alibaba Qwen | 1 474 | 1 457–1 491 | 1 295 | Proprietary |
| 38 | MiMo-V2 Pro | Xiaomi | 1 474 | 1 466–1 481 | 7 182 | Proprietary |
| 39 | GPT-5.6 Sol в режиме xhigh | OpenAI | 1 473 | 1 466–1 481 | 7 356 | Proprietary |
| 40 | Claude Sonnet 4.5 в режиме high | Anthropic | 1 472 | 1 467–1 477 | 20 389 | Proprietary |
| 41 | Claude Sonnet 4.5 | Anthropic | 1 471 | 1 466–1 476 | 20 548 | Proprietary |
| 42 | Kimi K2.6 | Moonshot AI | 1 470 | 1 463–1 476 | 11 742 | Modified MIT |
| 43 | Claude Sonnet 5 в режиме high | Anthropic | 1 467 | 1 460–1 474 | 10 085 | Proprietary |
| 44 | Grok 4.5 | xAI | 1 466 | 1 458–1 473 | 8 207 | Proprietary |
| 45 | DeepSeek V4 Pro 0423 | DeepSeek | 1 465 | 1 459–1 471 | 16 803 | MIT |
| 46 | GPT-5.4 | OpenAI | 1 464 | 1 459–1 470 | 19 823 | Proprietary |
| 47 | GPT-5.6 Terra в режиме xhigh | OpenAI | 1 464 | 1 456–1 471 | 7 573 | Proprietary |
| 48 | GPT-6 Astra в режиме max | OpenAI | 1 463 | 1 440–1 485 | 714 | Proprietary |
| 49 | Gemini 3 Flash Preview | 1 463 | 1 455–1 470 | 6 402 | Proprietary | |
| 50 | GPT-5.1 в режиме high | OpenAI | 1 462 | 1 455–1 469 | 8 555 | Proprietary |
| 51 | MiMo-V2.5 | Xiaomi | 1 461 | 1 455–1 468 | 14 189 | MIT |
| 52 | GLM 5 | Z.ai | 1 461 | 1 454–1 468 | 7 870 | MIT |
| 53 | Qwen3.7 Plus | Alibaba Qwen | 1 461 | 1 454–1 467 | 11 353 | Proprietary |
| 54 | Kimi K2.5 · thinking | Moonshot AI | 1 460 | 1 455–1 466 | 19 897 | Modified MIT |
| 55 | LongCat-Flash-Chat · chat-2602-exp | Meituan | 1 460 | 1 453–1 467 | 8 710 | Proprietary |
| 56 | DeepSeek V4 Pro 0423 в режиме high | DeepSeek | 1 459 | 1 453–1 466 | 15 887 | MIT |
| 57 | Amazon Nova Experimental Chat · 26-02-10 | Amazon | 1 459 | 1 437–1 480 | 704 | Proprietary |
| 58 | DeepSeek V4 Pro 0813 в режиме high | DeepSeek | 1 457 | 1 445–1 470 | 2 316 | MIT |
| 59 | GPT-5.6 Luna в режиме xhigh | OpenAI | 1 457 | 1 449–1 464 | 7 861 | Proprietary |
| 60 | MiniMax M3 | MiniMax | 1 457 | 1 451–1 463 | 14 203 | MiniMax Community License |
| 61 | Qwen3.6 Plus | Alibaba Qwen | 1 456 | 1 450–1 462 | 14 485 | Proprietary |
| 62 | GLM 5V Turbo | Z.ai | 1 456 | 1 444–1 467 | 2 641 | Proprietary |
| 63 | Qwen3.6 Max Preview | Alibaba Qwen | 1 455 | 1 441–1 470 | 1 638 | Proprietary |
| 64 | Gemma 4 31B | 1 455 | 1 439–1 471 | 1 293 | Apache 2.0 | |
| 65 | Grok 4.20 · beta-0309-reasoning | xAI | 1 455 | 1 449–1 460 | 19 401 | Proprietary |
| 66 | Qwen3.8 27B | Alibaba Qwen | 1 454 | 1 443–1 466 | 2 817 | Apache 2.0 |
| 67 | MiMo-V2 Omni | Xiaomi | 1 454 | 1 446–1 462 | 6 279 | Proprietary |
| 68 | Nemotron 3 Ultra | NVIDIA | 1 453 | 1 443–1 464 | 3 416 | OpenMDW-1.1 |
| 69 | Gemini 2.5 Pro | 1 453 | 1 449–1 457 | 28 292 | Proprietary | |
| 70 | Amazon Nova Experimental Chat · 12-10 | Amazon | 1 452 | 1 432–1 473 | 767 | Proprietary |
| 71 | Kimi K2.5 · instant | Moonshot AI | 1 452 | 1 438–1 466 | 1 724 | Modified MIT |
| 72 | Hy3 | Tencent | 1 452 | 1 439–1 465 | 2 085 | Apache 2.0 |
| 73 | GLM 4.7 | Z.ai | 1 451 | 1 440–1 462 | 2 617 | MIT |
| 74 | Claude Opus 4.1 · 20250805-thinking-16k | Anthropic | 1 451 | 1 445–1 457 | 10 525 | Proprietary |
| 75 | Qwen3.5 397B A17B | Alibaba Qwen | 1 450 | 1 445–1 455 | 23 429 | Apache 2.0 |
| 76 | GPT-5.2 Chat | OpenAI | 1 450 | 1 443–1 456 | 9 907 | Proprietary |
| 77 | Seed 2.0 Pro | ByteDance Seed | 1 450 | 1 444–1 455 | 22 044 | Proprietary |
| 78 | DeepSeek V4 Flash 0423 | DeepSeek | 1 449 | 1 443–1 455 | 15 297 | MIT |
| 79 | Gemma 4 26B A4B | 1 449 | 1 433–1 465 | 1 223 | Apache 2.0 | |
| 80 | Qwen3 Max · preview | Alibaba Qwen | 1 449 | 1 441–1 457 | 5 785 | Proprietary |
| 81 | Claude Opus 4.1 · 20250805 | Anthropic | 1 448 | 1 442–1 453 | 16 152 | Proprietary |
| 82 | ERNIE 5.0 · 0110 | Baidu | 1 448 | 1 441–1 454 | 8 798 | Proprietary |
| 83 | Inkling | Thinking Machines Lab | 1 446 | 1 439–1 454 | 7 143 | Apache 2.0 |
| 84 | Grok 4.20 Multi-Agent | xAI | 1 446 | 1 441–1 452 | 18 841 | Proprietary |
| 85 | DeepSeek V3.2 · v3.2-thinking | DeepSeek | 1 446 | 1 439–1 453 | 9 286 | MIT |
| 86 | Mistral Medium 3.5 | Mistral AI | 1 446 | 1 435–1 456 | 3 620 | Modified MIT |
| 87 | DeepSeek V3.2 Exp · exp-thinking | DeepSeek | 1 445 | 1 432–1 459 | 2 040 | MIT |
| 88 | LongCat-Flash-Chat · chat | Meituan | 1 445 | 1 433–1 457 | 2 412 | MIT |
| 89 | Grok 4.6 в режиме high | xAI | 1 445 | 1 435–1 455 | 4 230 | Proprietary |
| 90 | GLM 4.6 | Z.ai | 1 445 | 1 438–1 452 | 8 064 | MIT |
| 91 | DeepSeek V3.2 · v3.2 | DeepSeek | 1 445 | 1 438–1 451 | 10 731 | MIT |
| 92 | Grok 4.20 · beta1 | xAI | 1 444 | 1 437–1 452 | 8 029 | Proprietary |
| 93 | Gemini 3.5 Flash Lite | 1 442 | 1 434–1 450 | 7 235 | Proprietary | |
| 94 | Grok 3 | xAI | 1 441 | 1 433–1 450 | 5 819 | Proprietary |
| 95 | Grok 4.1 · 4.1 | xAI | 1 441 | 1 436–1 447 | 16 462 | Proprietary |
| 96 | MiniMax M2.7 | MiniMax | 1 440 | 1 435–1 446 | 19 931 | Modified MIT |
| 97 | Hunyuan Vision 1.5 | Tencent | 1 439 | 1 414–1 465 | 490 | Proprietary |
| 98 | ERNIE 5.0 · preview-1203 | Baidu | 1 439 | 1 426–1 452 | 2 106 | Proprietary |
| 99 | GPT-5.1 | OpenAI | 1 439 | 1 432–1 446 | 9 301 | Proprietary |
| 100 | DeepSeek V4 Flash 0423 в режиме high | DeepSeek | 1 439 | 1 433–1 445 | 15 097 | MIT |
| 101 | Qwen3 VL 235B A22B Instruct | Alibaba Qwen | 1 439 | 1 426–1 451 | 2 474 | Apache 2.0 |
| 102 | Qwen3 235B A22B Instruct 2507 | Alibaba Qwen | 1 437 | 1 432–1 441 | 22 145 | Apache 2.0 |
| 103 | Mistral Large 3 2512 | Mistral AI | 1 436 | 1 431–1 441 | 17 575 | Apache 2.0 |
| 104 | Amazon Nova Experimental Chat · 26-01-10 | Amazon | 1 436 | 1 415–1 457 | 719 | Proprietary |
| 105 | DeepSeek V3.2 Exp · exp | DeepSeek | 1 436 | 1 425–1 447 | 2 698 | MIT |
| 106 | Grok 4.1 · 4.1-thinking | xAI | 1 436 | 1 430–1 441 | 16 063 | Proprietary |
| 107 | MiMo-V2 Flash · flash (non-thinking) | Xiaomi | 1 435 | 1 429–1 441 | 11 793 | MIT |
| 108 | Mistral Medium 3.1 | Mistral AI | 1 435 | 1 430–1 439 | 22 553 | Proprietary |
| 109 | Claude Haiku 4.5 | Anthropic | 1 435 | 1 430–1 439 | 35 467 | Proprietary |
| 110 | DeepSeek V3.1 Terminus · terminus-thinking | DeepSeek | 1 434 | 1 412–1 455 | 722 | MIT |
| 111 | Gemini 3 Flash Preview в режиме minimal | 1 433 | 1 428–1 438 | 24 805 | Proprietary | |
| 112 | Qwen3.5-122B-A10B | Alibaba Qwen | 1 433 | 1 426–1 440 | 8 269 | Apache 2.0 |
| 113 | Kimi K2 Thinking | Moonshot AI | 1 433 | 1 428–1 438 | 15 038 | Modified MIT |
| 114 | Qwen3 235B A22B Thinking 2507 | Alibaba Qwen | 1 431 | 1 417–1 446 | 1 716 | Apache 2.0 |
| 115 | Qwen3 Next 80B A3B Instruct | Alibaba Qwen | 1 431 | 1 423–1 440 | 5 140 | Apache 2.0 |
| 116 | GPT-5.2 в режиме high | OpenAI | 1 430 | 1 424–1 436 | 12 509 | Proprietary |
| 117 | GLM 4.5 | Z.ai | 1 429 | 1 420–1 438 | 4 800 | MIT |
| 118 | Qwen3 Max · 2025-09-23 | Alibaba Qwen | 1 429 | 1 416–1 442 | 2 097 | Proprietary |
| 119 | ChatGPT-4o (latest) | OpenAI | 1 429 | 1 424–1 434 | 17 046 | Proprietary |
| 120 | Inkling Small | Thinking Machines Lab | 1 429 | 1 420–1 437 | 5 196 | Apache 2.0 |
| 121 | DeepSeek V3.1 · v3.1-thinking | DeepSeek | 1 427 | 1 414–1 440 | 2 133 | MIT |
| 122 | Qwen3.5-27B | Alibaba Qwen | 1 427 | 1 419–1 434 | 7 959 | Apache 2.0 |
| 123 | Amazon Nova Experimental Chat · 11-10 | Amazon | 1 425 | 1 417–1 433 | 5 438 | Proprietary |
| 124 | Solar Pro 4 | Upstage | 1 425 | 1 401–1 448 | 642 | Proprietary |
| 125 | GPT-5.5 · 5.5-instant | OpenAI | 1 424 | 1 416–1 431 | 8 742 | Proprietary |
| 126 | GPT-5.2 | OpenAI | 1 423 | 1 418–1 428 | 23 266 | Proprietary |
| 127 | Hy3 preview | Tencent | 1 422 | 1 409–1 435 | 2 226 | tencent-hunyuan-community |
| 128 | R1 0528 | DeepSeek | 1 422 | 1 411–1 433 | 3 096 | MIT |
| 129 | Step 3.5 Flash | StepFun | 1 422 | 1 416–1 427 | 15 418 | Apache 2.0 |
| 130 | GPT-5.4 Mini в режиме high | OpenAI | 1 421 | 1 415–1 426 | 18 715 | Proprietary |
| 131 | Gemini 2.5 Flash · flash | 1 421 | 1 416–1 425 | 28 236 | Proprietary | |
| 132 | DeepSeek V3.1 · v3.1 | DeepSeek | 1 420 | 1 409–1 431 | 2 875 | MIT |
| 133 | GPT-5 в режиме high | OpenAI | 1 420 | 1 412–1 428 | 6 583 | Proprietary |
| 134 | MiniMax M2.1 | MiniMax | 1 419 | 1 409–1 429 | 3 425 | MIT |
| 135 | MiMo-V2 Flash · flash (thinking) | Xiaomi | 1 418 | 1 406–1 430 | 2 276 | MIT |
| 136 | Grok 4 Fast · chat | xAI | 1 417 | 1 400–1 433 | 1 271 | Proprietary |
| 137 | Qwen3 VL 235B A22B Thinking | Alibaba Qwen | 1 417 | 1 403–1 430 | 1 805 | Apache 2.0 |
| 138 | Grok 4.1 Fast | xAI | 1 415 | 1 409–1 421 | 14 088 | Proprietary |
| 139 | ERNIE 5.0 · preview-1022 | Baidu | 1 414 | 1 398–1 430 | 1 211 | Proprietary |
| 140 | Amazon Nova Experimental Chat · 10-20 | Amazon | 1 414 | 1 403–1 426 | 2 575 | Proprietary |
| 141 | Grok 4 Fast · reasoning | xAI | 1 412 | 1 403–1 421 | 4 324 | Proprietary |
| 142 | Gemini 2.5 Flash · flash-preview-09-2025 | 1 411 | 1 404–1 419 | 7 395 | Proprietary | |
| 143 | o3 | OpenAI | 1 411 | 1 405–1 417 | 12 039 | Proprietary |
| 144 | Grok 4 | xAI | 1 410 | 1 403–1 417 | 8 700 | Proprietary |
| 145 | Qwen3.5-35B-A3B | Alibaba Qwen | 1 410 | 1 403–1 417 | 8 463 | Apache 2.0 |
| 146 | Qwen3.5-Flash | Alibaba Qwen | 1 409 | 1 403–1 415 | 17 191 | Proprietary |
| 147 | DeepSeek V3.1 Terminus · terminus | DeepSeek | 1 407 | 1 387–1 427 | 851 | MIT |
| 148 | GPT-4.5 Preview | OpenAI | 1 407 | 1 395–1 419 | 2 260 | Proprietary |
| 149 | Qwen3 30B A3B Instruct 2507 | Alibaba Qwen | 1 406 | 1 398–1 415 | 4 729 | Apache 2.0 |
| 150 | GPT-5 | OpenAI | 1 405 | 1 398–1 413 | 6 552 | Proprietary |
| 151 | Claude Opus 4 · 20250514-thinking-16k | Anthropic | 1 404 | 1 396–1 411 | 7 150 | Proprietary |
| 152 | Gemini 3.1 Flash Lite Preview | 1 403 | 1 397–1 408 | 18 823 | Proprietary | |
| 153 | Grok 4.3 | xAI | 1 401 | 1 395–1 406 | 21 085 | Proprietary |
| 154 | Muse Glimmer 30B | Meta | 1 399 | 1 381–1 418 | 987 | Apache-2.0 |
| 155 | Hunyuan T1 | Tencent | 1 398 | 1 377–1 418 | 788 | Proprietary |
| 156 | Nemotron 3 Super | NVIDIA | 1 398 | 1 384–1 412 | 1 698 | NVIDIA Open Model |
| 157 | GPT-5.3 Chat | OpenAI | 1 395 | 1 388–1 402 | 9 403 | Proprietary |
| 158 | Qwen3 235B A22B · a22b-no-thinking | Alibaba Qwen | 1 393 | 1 385–1 400 | 7 367 | Apache 2.0 |
| 159 | GPT-4.1 | OpenAI | 1 393 | 1 386–1 399 | 10 328 | Proprietary |
| 160 | GLM 4.6V | Z.ai | 1 390 | 1 367–1 414 | 624 | MIT |
| 161 | Granite 4.2 30B | IBM Granite | 1 390 | 1 370–1 410 | 914 | Apache 2.0 |
| 162 | Ling-flash-2.0 | inclusionAI | 1 388 | 1 374–1 403 | 1 574 | MIT |
| 163 | GLM 4.5 Air | Z.ai | 1 388 | 1 381–1 396 | 6 490 | MIT |
| 164 | GPT-5.4 Nano в режиме high | OpenAI | 1 388 | 1 382–1 393 | 18 796 | Proprietary |
| 165 | Nova 2 Lite | Amazon | 1 386 | 1 375–1 398 | 2 682 | Proprietary |
| 166 | Qwen3 Next 80B A3B Thinking | Alibaba Qwen | 1 385 | 1 374–1 396 | 2 995 | Apache 2.0 |
| 167 | Claude Sonnet 4 · 20250514-thinking-32k | Anthropic | 1 384 | 1 377–1 392 | 6 927 | Proprietary |
| 168 | Gemini 2.5 Flash Lite · 09-2025-no-thinking | 1 384 | 1 377–1 390 | 10 412 | Proprietary | |
| 169 | MiniMax M2.5 | MiniMax | 1 384 | 1 377–1 390 | 11 662 | Modified MIT |
| 170 | GPT-5 Mini в режиме high | OpenAI | 1 382 | 1 374–1 391 | 5 513 | Proprietary |
| 171 | Claude Opus 4 · 20250514 | Anthropic | 1 382 | 1 375–1 389 | 8 462 | Proprietary |
| 172 | Qwen3 Coder 480B A35B | Alibaba Qwen | 1 382 | 1 373–1 390 | 5 115 | Apache 2.0 |
| 173 | DeepSeek V3 0324 | DeepSeek | 1 380 | 1 373–1 386 | 8 978 | MIT |
| 174 | Step 3 | StepFun | 1 379 | 1 363–1 395 | 1 326 | Apache 2.0 |
| 175 | Kimi K2 0905 | Moonshot AI | 1 378 | 1 366–1 390 | 2 377 | Modified MIT |
| 176 | MiniMax M2 | MiniMax | 1 378 | 1 363–1 392 | 1 609 | Apache 2.0 |
| 177 | Grok 3 Mini в режиме high | xAI | 1 378 | 1 368–1 388 | 3 367 | Proprietary |
| 178 | Mercury 2 | Inception Labs | 1 377 | 1 355–1 398 | 702 | Proprietary |
| 179 | Nemotron 3 Nano 30B A3B | NVIDIA | 1 377 | 1 366–1 387 | 3 269 | NVIDIA Open Model |
| 180 | Ring-flash-2.0 | inclusionAI | 1 376 | 1 362–1 391 | 1 564 | MIT |
| 181 | Hunyuan TurboS · 20250416 | Tencent | 1 376 | 1 363–1 389 | 2 045 | Proprietary |
| 182 | o1 · preview | OpenAI | 1 376 | 1 366–1 385 | 4 917 | Proprietary |
| 183 | o3 Mini High | OpenAI | 1 376 | 1 365–1 387 | 2 944 | Proprietary |
| 184 | Gemini 2.5 Flash Lite · 06-17-thinking | 1 376 | 1 368–1 383 | 6 422 | Proprietary | |
| 185 | Grok 3 Mini | xAI | 1 375 | 1 366–1 384 | 4 466 | Proprietary |
| 186 | R1 | DeepSeek | 1 375 | 1 364–1 386 | 2 656 | MIT |
| 187 | Nemotron 3.5 Lightning | NVIDIA | 1 375 | 1 362–1 388 | 2 232 | OpenMDW-1.1 |
| 188 | Mistral Medium 3 | Mistral AI | 1 374 | 1 366–1 382 | 6 353 | Proprietary |
| 189 | INTELLECT-3 | Prime Intellect | 1 374 | 1 357–1 390 | 1 247 | MIT |
| 190 | o1 · 2024-12-17 | OpenAI | 1 372 | 1 363–1 382 | 4 238 | Proprietary |
| 191 | Qwen3 235B A22B · a22b | Alibaba Qwen | 1 372 | 1 363–1 380 | 5 085 | Apache 2.0 |
| 192 | Kimi K2 0711 | Moonshot AI | 1 370 | 1 361–1 378 | 5 371 | Modified MIT |
| 193 | gpt-oss-120b | OpenAI | 1 369 | 1 361–1 377 | 6 402 | Apache 2.0 |
| 194 | GLM 4.7 Flash | Z.ai | 1 368 | 1 356–1 379 | 2 465 | MIT |
| 195 | Trinity Large | Arcee AI | 1 367 | 1 360–1 375 | 8 603 | Apache 2.0 |
| 196 | o4 Mini | OpenAI | 1 367 | 1 360–1 373 | 9 056 | Proprietary |
| 197 | Amazon Nova Experimental Chat · 10-09 | Amazon | 1 366 | 1 344–1 387 | 673 | Proprietary |
| 198 | Qwen2.5 Max | Alibaba Qwen | 1 362 | 1 354–1 370 | 5 690 | Proprietary |
| 199 | GPT-4.1 Mini | OpenAI | 1 362 | 1 355–1 369 | 7 659 | Proprietary |
| 200 | GLM 4.5V | Z.ai | 1 359 | 1 341–1 377 | 1 011 | MIT |
| 201 | Claude Sonnet 4 · 20250514 | Anthropic | 1 359 | 1 352–1 366 | 7 861 | Proprietary |
| 202 | Mistral Small 3.2 24B | Mistral AI | 1 358 | 1 348–1 368 | 3 432 | Apache 2.0 |
| 203 | Trinity Large Thinking | Arcee AI | 1 356 | 1 349–1 364 | 9 491 | Apache 2.0 |
| 204 | Gemini 2.0 Flash | 1 356 | 1 349–1 363 | 7 914 | Proprietary | |
| 205 | Step-1o Turbo | StepFun | 1 354 | 1 340–1 368 | 1 696 | Proprietary |
| 206 | MiniMax M1 | MiniMax | 1 354 | 1 346–1 361 | 6 942 | Apache 2.0 |
| 207 | Gemma 3 27B | 1 354 | 1 347–1 360 | 8 889 | Gemma | |
| 208 | o1-mini | OpenAI | 1 351 | 1 344–1 359 | 8 369 | Proprietary |
| 209 | OLMo 3.1 32B Instruct | Ai2 | 1 350 | 1 338–1 363 | 2 346 | Apache 2.0 |
| 210 | Qwen3 32B | Alibaba Qwen | 1 350 | 1 329–1 370 | 729 | Apache 2.0 |
| 211 | o3 Mini | OpenAI | 1 343 | 1 337–1 349 | 10 227 | Proprietary |
| 212 | QwQ 32B | Alibaba Qwen | 1 341 | 1 333–1 350 | 4 633 | Apache 2.0 |
| 213 | Claude 3.7 Sonnet · 20250219-thinking-32k | Anthropic | 1 340 | 1 333–1 348 | 6 958 | Proprietary |
| 214 | Llama 3.3 Nemotron Super 49B v1.5 | NVIDIA | 1 340 | 1 317–1 363 | 608 | Nvidia Open |
| 215 | GPT-5 Nano в режиме high | OpenAI | 1 339 | 1 324–1 353 | 1 696 | Proprietary |
| 216 | Granite 4.2 3B | IBM Granite | 1 338 | 1 316–1 360 | 809 | Apache 2.0 |
| 217 | Granite 4.2 8B | IBM Granite | 1 338 | 1 317–1 360 | 861 | Apache 2.0 |
| 218 | Llama 3.3 Nemotron Super 49B v1 | NVIDIA | 1 338 | 1 309–1 367 | 344 | Nvidia |
| 219 | Hunyuan TurboS · 20250226 | Tencent | 1 337 | 1 310–1 365 | 367 | Proprietary |
| 220 | Llama 3.1 Nemotron Ultra 253B v1 | NVIDIA | 1 336 | 1 310–1 363 | 448 | Nvidia Open Model |
| 221 | Command A | Cohere | 1 336 | 1 330–1 342 | 11 170 | CC-BY-NC-4.0 |
| 222 | OLMo 3 32B Think | Ai2 | 1 336 | 1 319–1 352 | 1 253 | Apache 2.0 |
| 223 | Qwen-Plus | Alibaba Qwen | 1 333 | 1 316–1 350 | 964 | Proprietary |
| 224 | Gemini 2.0 Flash-Lite | 1 330 | 1 321–1 339 | 4 113 | Proprietary | |
| 225 | Qwen3 30B A3B | Alibaba Qwen | 1 329 | 1 321–1 338 | 5 143 | Apache 2.0 |
| 226 | Hunyuan Turbo | Tencent | 1 327 | 1 300–1 354 | 343 | Proprietary |
| 227 | OLMo 3.1 32B Think | Ai2 | 1 326 | 1 312–1 341 | 1 671 | Apache 2.0 |
| 228 | DeepSeek V3 | DeepSeek | 1 324 | 1 314–1 333 | 3 637 | DeepSeek |
| 229 | Yi-Lightning | 01.AI | 1 322 | 1 312–1 332 | 3 939 | Proprietary |
| 230 | Claude 3.7 Sonnet · 20250219 | Anthropic | 1 320 | 1 313–1 327 | 7 954 | Proprietary |
| 231 | Gemma 3 12B | 1 318 | 1 298–1 339 | 700 | Gemma | |
| 232 | Qwen2.5 Plus | Alibaba Qwen | 1 318 | 1 305–1 331 | 1 713 | Proprietary |
| 233 | Granite 4.1 8B | IBM Granite | 1 314 | 1 296–1 331 | 1 351 | Apache 2.0 |
| 234 | Claude 3.5 Sonnet · 20241022 | Anthropic | 1 311 | 1 306–1 317 | 15 025 | Proprietary |
| 235 | Mercury | Inception Labs | 1 310 | 1 282–1 338 | 441 | Proprietary |
| 236 | Step-2 16k | StepFun | 1 309 | 1 290–1 327 | 786 | Proprietary |
| 237 | Athene V2 Chat | Nexusflow | 1 304 | 1 295–1 313 | 4 039 | NexusFlow |
| 238 | DeepSeek V2.5 · v2.5-1210 | DeepSeek | 1 299 | 1 283–1 315 | 1 132 | DeepSeek |
| 239 | Gemini 1.5 Pro · 1.5-pro-002 | 1 299 | 1 292–1 306 | 9 056 | Proprietary | |
| 240 | Llama 4 Maverick | Meta | 1 296 | 1 288–1 303 | 7 675 | Llama 4 |
| 241 | Hunyuan Large | Tencent | 1 295 | 1 272–1 317 | 601 | Proprietary |
| 242 | GPT-4o (2024-05-13) | OpenAI | 1 293 | 1 286–1 299 | 19 579 | Proprietary |
| 243 | GPT-4.1 Nano | OpenAI | 1 293 | 1 276–1 310 | 1 062 | Proprietary |
| 244 | Llama 4 Scout | Meta | 1 293 | 1 284–1 301 | 5 716 | Llama |
| 245 | gpt-oss-20b | OpenAI | 1 292 | 1 278–1 305 | 2 083 | Apache 2.0 |
| 246 | GLM-4-Plus · plus-0111 | Z.ai | 1 292 | 1 274–1 309 | 955 | Proprietary |
| 247 | Llama 3.1 405B Instruct · bf16 | Meta | 1 292 | 1 284–1 299 | 6 791 | Llama 3.1 Community |
| 248 | Gemma 3n E4B | 1 292 | 1 282–1 301 | 4 104 | Gemma | |
| 249 | Mistral Small 3.1 24B | Mistral AI | 1 290 | 1 282–1 298 | 6 489 | Apache 2.0 |
| 250 | Magistral Medium | Mistral AI | 1 285 | 1 272–1 297 | 2 386 | Proprietary |
| 251 | Llama 3.1 405B Instruct · fp8 | Meta | 1 283 | 1 276–1 290 | 9 964 | Llama 3.1 Community |
| 252 | Llama 3.1 Nemotron 70B Instruct | NVIDIA | 1 283 | 1 266–1 299 | 1 136 | Llama 3.1 |
| 253 | Claude 3.5 Sonnet · 20240620 | Anthropic | 1 283 | 1 276–1 289 | 14 269 | Proprietary |
| 254 | Qwen Max | Alibaba Qwen | 1 281 | 1 269–1 292 | 2 531 | Qwen |
| 255 | Grok 2 | xAI | 1 280 | 1 273–1 287 | 10 486 | Proprietary |
| 256 | DeepSeek V2.5 · v2.5 | DeepSeek | 1 280 | 1 271–1 289 | 4 055 | DeepSeek |
| 257 | GLM-4-Plus · plus | Z.ai | 1 280 | 1 270–1 289 | 4 016 | Proprietary |
| 258 | Qwen2.5 72B Instruct | Alibaba Qwen | 1 280 | 1 272–1 287 | 6 208 | Qwen |
| 259 | GPT-4o-mini (2024-07-18) | OpenAI | 1 279 | 1 272–1 285 | 11 352 | Proprietary |
| 260 | Llama 3.3 70B Instruct | Meta | 1 277 | 1 271–1 283 | 9 527 | Llama-3.3 |
| 261 | GPT-4o (2024-08-06) | OpenAI | 1 277 | 1 269–1 285 | 8 010 | Proprietary |
| 262 | Hunyuan Standard · 2025-02-10 | Tencent | 1 274 | 1 251–1 297 | 587 | Proprietary |
| 263 | Hunyuan Large Vision | Tencent | 1 272 | 1 253–1 290 | 1 058 | Proprietary |
| 264 | Qwen2.5 Coder 32B Instruct | Alibaba Qwen | 1 269 | 1 250–1 289 | 759 | Apache 2.0 |
| 265 | Mistral Large 2407 | Mistral AI | 1 269 | 1 262–1 277 | 7 818 | Mistral Research |
| 266 | Mistral Large | Mistral AI | 1 268 | 1 259–1 276 | 4 452 | MRL |
| 267 | Grok 2 Mini | xAI | 1 268 | 1 261–1 275 | 8 664 | Proprietary |
| 268 | Granite 4.0 H Small | IBM Granite | 1 267 | 1 250–1 284 | 1 343 | Apache 2.0 |
| 269 | GPT-4 Turbo | OpenAI | 1 266 | 1 259–1 273 | 17 922 | Proprietary |
| 270 | Gemini 1.5 Flash · 002 | 1 265 | 1 257–1 273 | 5 484 | Proprietary | |
| 271 | Claude 3.5 Haiku | Anthropic | 1 264 | 1 258–1 270 | 12 309 | Proprietary |
| 272 | Llama 3.1 70B Instruct | Meta | 1 261 | 1 254–1 269 | 9 133 | Llama 3.1 Community |
| 273 | Gemini 1.5 Pro · 1.5-pro-001 | 1 259 | 1 252–1 266 | 13 552 | Proprietary | |
| 274 | Nova Pro 1.0 | Amazon | 1 259 | 1 250–1 268 | 4 069 | Proprietary |
| 275 | Athene 70B | Nexusflow | 1 255 | 1 245–1 265 | 3 605 | CC-BY-NC-4.0 |
| 276 | GPT-4 Turbo Preview · 1106-preview | OpenAI | 1 254 | 1 247–1 261 | 18 316 | Proprietary |
| 277 | Gemma 3 4B | 1 253 | 1 233–1 274 | 764 | Gemma | |
| 278 | GPT-4 Turbo Preview · 0125-preview | OpenAI | 1 251 | 1 244–1 259 | 16 792 | Proprietary |
| 279 | Gemini 1.5 Pro · advanced-0514 | 1 250 | 1 241–1 258 | 8 405 | Proprietary | |
| 280 | OLMo 2 32B Instruct | Ai2 | 1 244 | 1 220–1 267 | 538 | Apache-2.0 |
| 281 | Claude 3 Opus | Anthropic | 1 243 | 1 238–1 249 | 34 293 | Proprietary |
| 282 | Mistral Small 3 | Mistral AI | 1 243 | 1 232–1 255 | 2 355 | Apache 2.0 |
| 283 | Llama 3.1 Tulu 3 70B | Ai2 | 1 237 | 1 213–1 260 | 499 | Llama 3.1 |
| 284 | Jamba 1.5 Large | AI21 Labs | 1 231 | 1 217–1 245 | 1 528 | Jamba Open |
| 285 | Llama 3 70B Instruct | Meta | 1 231 | 1 224–1 237 | 29 608 | Llama 3 Community |
| 286 | Phi 4 | Microsoft | 1 230 | 1 221–1 239 | 3 804 | MIT |
| 287 | Hunyuan Standard · 256k | Tencent | 1 228 | 1 200–1 256 | 370 | Proprietary |
| 288 | Nova Lite 1.0 | Amazon | 1 226 | 1 216–1 236 | 3 136 | Proprietary |
| 289 | Gemini 1.5 Flash · 001 | 1 225 | 1 217–1 232 | 10 974 | Proprietary | |
| 290 | GLM-4 | Z.ai | 1 219 | 1 205–1 232 | 1 775 | Proprietary |
| 291 | DeepSeek Coder V2 | DeepSeek | 1 218 | 1 206–1 229 | 2 699 | DeepSeek License |
| 292 | Reka Core | Reka AI | 1 216 | 1 202–1 230 | 1 341 | Proprietary |
| 293 | Gemini 1.5 Flash-8B | 1 215 | 1 207–1 223 | 5 581 | Proprietary | |
| 294 | GPT-4 | OpenAI | 1 209 | 1 200–1 218 | 9 838 | Proprietary |
| 295 | Ministral 8B (2410) | Mistral AI | 1 206 | 1 186–1 226 | 703 | MRL |
| 296 | Gemma 2 27B | 1 206 | 1 200–1 212 | 12 606 | Gemma license | |
| 297 | Qwen2 72B Instruct | Alibaba Qwen | 1 206 | 1 197–1 215 | 6 483 | Qianwen LICENSE |
| 298 | Llama 3.1 Nemotron 51B Instruct | NVIDIA | 1 205 | 1 183–1 228 | 578 | Llama 3.1 |
| 299 | Nova Micro 1.0 | Amazon | 1 203 | 1 193–1 213 | 3 050 | Proprietary |
| 300 | Claude 3 Sonnet | Anthropic | 1 203 | 1 195–1 210 | 19 193 | Proprietary |
Данные на 13 сентября 2026. Как мы считаем: методика
Свежесть данных
Снимок замеров опубликован 13 сентября 2026; всего снимков борда в архиве: 200.
- замеры оценок людей: загружено 18 сентября 2026, 08:25 UTC
- каталог моделей и цены: загружено 18 сентября 2026, 08:51 UTC
- доли использования: загружено 18 сентября 2026, 08:51 UTC
Другие рейтинги
- Рейтинг агентов
- Рейтинг работы с документами
- Рейтинг редактирования изображений
- Рейтинг видео из изображения
- Популярность моделей по использованию через API-агрегаторы
- Рейтинг поисковых моделей
- Рейтинг моделей для кода
- Рейтинг моделей для творческих текстов
- Рейтинг моделей для математики
- Рейтинг моделей для русского языка
- Рейтинг генерации изображений
- Рейтинг генерации видео
- Рейтинг редактирования видео
- Рейтинг моделей с изображениями на входе
- Рейтинг моделей для веб-разработки
Данные и источники
Обновлено 18 сентября 2026. Каталог, цены и режимы проверяются ежедневно по нескольким API-агрегаторам и провайдерам. Места — по открытым замерам оценок людей, интервалы приведены как в замере.