AI 模型速度排行榜

按实测输出吞吐(tokens/秒)与平均延迟排名。

  1. Z.ai GLM 5.2 — 吞吐: 25000.0 tps, 延迟: 889.00s
  2. xAI Grok 4.20 Multi-Agent — 吞吐: 5082.4 tps, 延迟: 3320.00s
  3. ByteDance Seedance 2.0 Mini — 吞吐: 1774.0 tps, 延迟: 0.00s
  4. ByteDance Seedance 2.0 Fast — 吞吐: 782.0 tps, 延迟: 139251.00s
  5. ByteDance Doubao-Seedream-5.0-pro — 吞吐: 523.5 tps, 延迟: 30994.00s
  6. Google Gemini 3.6 Flash — 吞吐: 461.8 tps, 延迟: 6717.25s
  7. ByteDance Seedance 2.0 — 吞吐: 429.5 tps, 延迟: 253535.00s
  8. Anthropic Claude Claude Fable 5.1 — 吞吐: 386.8 tps, 延迟: 25555.64s
  9. Google Gemini 3.5 Flash — 吞吐: 279.5 tps, 延迟: 3363.11s
  10. Google Gemini 3.1 Pro Preview — 吞吐: 257.1 tps, 延迟: 14217.74s
  11. Google Gemini 3 Flash Preview — 吞吐: 229.8 tps, 延迟: 2178.82s
  12. MiniMax MiniMax M3 — 吞吐: 226.4 tps, 延迟: 0.00s
  13. OpenAI dall-e-2 — 吞吐: 224.7 tps, 延迟: 18577.00s
  14. Google Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image 🍌) — 吞吐: 212.9 tps, 延迟: 6888.85s
  15. OpenAI GPT-6 Sol — 吞吐: 212.8 tps, 延迟: 708.00s
  16. Anthropic Claude Claude Haiku 5.5 — 吞吐: 212.6 tps, 延迟: 1382.33s
  17. Google Gemini 3.8 Flash — 吞吐: 188.1 tps, 延迟: 5445.46s
  18. Google Gemini 3.1 Flash-Lite Preview — 吞吐: 186.1 tps, 延迟: 0.00s
  19. OpenAI GPT-5.2 Pro — 吞吐: 176.5 tps, 延迟: 3694.00s
  20. OpenAI GPT-5.6 Luna — 吞吐: 175.8 tps, 延迟: 1832.45s
  21. OpenAI GPT-5.4 Mini — 吞吐: 174.6 tps, 延迟: 9877.10s
  22. xAI Grok Build 0.1 — 吞吐: 163.7 tps, 延迟: 942.00s
  23. DeepSeek DeepSeek V4.1 Flash — 吞吐: 163.3 tps, 延迟: 2156.04s
  24. OpenAI GPT Image 1.5 — 吞吐: 159.9 tps, 延迟: 41353.50s
  25. OpenAI dall-e-3 — 吞吐: 145.0 tps, 延迟: 28784.00s
  26. OpenAI gpt-audio-mini — 吞吐: 143.7 tps, 延迟: 384.00s
  27. MiniMax minimax-m2.1 — 吞吐: 129.9 tps, 延迟: 1483.00s
  28. Google Gemini Nano Banana 2.1 🍌 — 吞吐: 126.0 tps, 延迟: 0.00s
  29. Google Gemini 3.7 Flash — 吞吐: 116.5 tps, 延迟: 4367.12s
  30. Google Gemini 3.5 Flash-Lite — 吞吐: 115.6 tps, 延迟: 2609.91s
  31. Meta Muse Wonder 1.0 — 吞吐: 114.3 tps, 延迟: 3689.24s
  32. OpenAI gpt-5-nano — 吞吐: 113.4 tps, 延迟: 4831.11s
  33. ByteDance Doubao-Seed-2.0-mini 260428 — 吞吐: 109.2 tps, 延迟: 0.00s
  34. OpenAI gpt-image-1-mini — 吞吐: 109.1 tps, 延迟: 14443.25s
  35. xAI grok-4-fast-reasoning — 吞吐: 109.1 tps, 延迟: 0.00s
  36. Google Nano Banana (Gemini 2.5 Flash Image 🍌) — 吞吐: 108.3 tps, 延迟: 10918.00s
  37. OpenAI GPT-5.4 Nano — 吞吐: 106.3 tps, 延迟: 3842.17s
  38. xAI Grok 4.20 — 吞吐: 103.4 tps, 延迟: 0.00s
  39. xAI grok-4-0709 — 吞吐: 102.0 tps, 延迟: 1530.50s
  40. OpenAI GPT-5.6 Terra — 吞吐: 102.0 tps, 延迟: 6841.57s
  41. xAI Grok 4.3 — 吞吐: 100.5 tps, 延迟: 1778.33s
  42. OpenAI GPT-5.5 — 吞吐: 99.8 tps, 延迟: 4199.56s
  43. Meta Muse Image 1.0 — 吞吐: 98.5 tps, 延迟: 16214.04s
  44. Z.ai GLM 5.3 — 吞吐: 92.4 tps, 延迟: 0.00s
  45. Anthropic Claude Claude Sonnet 5 — 吞吐: 90.2 tps, 延迟: 2816.74s
  46. Anthropic Claude Claude Sonnet 5.5 — 吞吐: 89.5 tps, 延迟: 2753.79s
  47. Thinking Machines Thinking Machines: Inkling — 吞吐: 88.0 tps, 延迟: 0.00s
  48. Anthropic Claude Claude Opus 5.5 — 吞吐: 87.3 tps, 延迟: 4338.23s
  49. OpenAI GPT Image 1.0 — 吞吐: 87.1 tps, 延迟: 71642.00s
  50. OpenAI o3-mini — 吞吐: 86.3 tps, 延迟: 0.00s
  51. xAI grok-3 — 吞吐: 83.3 tps, 延迟: 0.00s
  52. xAI grok-3-mini — 吞吐: 83.0 tps, 延迟: 0.00s
  53. OpenAI gpt-3.5-turbo — 吞吐: 82.3 tps, 延迟: 624.33s
  54. Z.ai GLM 5.3 Flash — 吞吐: 76.5 tps, 延迟: 728.25s
  55. xAI Grok 4.20 (Non-Reasoning) — 吞吐: 75.8 tps, 延迟: 0.00s
  56. OpenAI GPT Image 2.5 Flare — 吞吐: 73.6 tps, 延迟: 25592.65s
  57. OpenAI GPT-5.2 — 吞吐: 71.2 tps, 延迟: 554.00s
  58. Alibaba Qwen3.7 Flash — 吞吐: 69.6 tps, 延迟: 631.00s
  59. Anthropic Claude Claude Opus 5 — 吞吐: 69.1 tps, 延迟: 3805.33s
  60. Meta Muse Spark 1.3 — 吞吐: 68.7 tps, 延迟: 0.00s
  61. MiniMax MiniMax-M2.7 — 吞吐: 68.2 tps, 延迟: 1108.00s
  62. xAI grok-4-1-fast-reasoning — 吞吐: 61.4 tps, 延迟: 0.00s
  63. MiniMax MiniMax-M2.7 Highspeed — 吞吐: 61.1 tps, 延迟: 1666.00s
  64. Alibaba Qwen3.8-Flash — 吞吐: 59.9 tps, 延迟: 0.00s
  65. MiniMax MiniMax M2.5 — 吞吐: 59.8 tps, 延迟: 1193.00s
  66. Alibaba Qwen3.6 Plus — 吞吐: 59.8 tps, 延迟: 1795.00s
  67. Alibaba Qwen3.5 Plus 2026-02-15 — 吞吐: 58.8 tps, 延迟: 1480.00s
  68. OpenAI gpt-4o-transcribe-diarize — 吞吐: 58.1 tps, 延迟: 0.00s
  69. xAI Grok 4.6 — 吞吐: 57.4 tps, 延迟: 4333.00s
  70. OpenAI GPT-5.4 — 吞吐: 57.3 tps, 延迟: 1310.00s
  71. Google Gemini 4 Argon — 吞吐: 56.9 tps, 延迟: 5540.00s
  72. Alibaba Qwen3.7 Max — 吞吐: 55.5 tps, 延迟: 903.00s
  73. OpenAI gpt-5-mini — 吞吐: 55.3 tps, 延迟: 2775.41s
  74. Anthropic Claude Claude Opus 4.8 — 吞吐: 54.5 tps, 延迟: 1767.36s
  75. Alibaba Qwen3.8-Max — 吞吐: 53.1 tps, 延迟: 2531.00s
  76. Anthropic Claude Claude Opus 4.7 — 吞吐: 49.1 tps, 延迟: 3179.38s
  77. DeepSeek DeepSeek-V4-Pro-0813 — 吞吐: 49.0 tps, 延迟: 1396.28s
  78. DeepSeek DeepSeek-V4-Flash 0731 — 吞吐: 48.8 tps, 延迟: 622.00s
  79. OpenAI GPT Image 2.5 Sunburst — 吞吐: 48.2 tps, 延迟: 33622.16s
  80. xAI Grok 4.7 — 吞吐: 47.5 tps, 延迟: 1890.25s
  81. xAI Grok 4.5 — 吞吐: 46.4 tps, 延迟: 2394.67s
  82. OpenAI GPT-6 Luna — 吞吐: 46.1 tps, 延迟: 2511.70s
  83. ByteDance Doubao-Seed-2.0-mini — 吞吐: 45.7 tps, 延迟: 0.00s
  84. Google Gemini 3.1 Flash-Lite — 吞吐: 44.9 tps, 延迟: 0.00s
  85. OpenAI GPT-6.1 Sol — 吞吐: 42.4 tps, 延迟: 1109.83s
  86. Moonshot kimi-k2-thinking — 吞吐: 41.7 tps, 延迟: 671.00s
  87. MiniMax minimax-m2 — 吞吐: 40.3 tps, 延迟: 0.00s
  88. ByteDance Doubao-Seed-2.1-pro — 吞吐: 37.6 tps, 延迟: 0.00s
  89. OpenAI gpt-3.5-turbo-16k — 吞吐: 35.9 tps, 延迟: 0.00s
  90. Moonshot Kimi K2.6 — 吞吐: 35.7 tps, 延迟: 14874.09s
  91. MiniMax MiniMax M2.5 Lightning — 吞吐: 34.3 tps, 延迟: 0.00s
  92. OpenAI gpt-4o-transcribe — 吞吐: 33.9 tps, 延迟: 0.00s
  93. Alibaba qwen-plus-2025-07-14 — 吞吐: 32.9 tps, 延迟: 0.00s
  94. DeepSeek DeepSeek-V3.2 — 吞吐: 32.7 tps, 延迟: 1989.00s
  95. Google Nano Banana Pro (Gemini 3 Pro Image 🍌) — 吞吐: 32.6 tps, 延迟: 27227.07s
  96. MiniMax MiniMax-M2.1-lightning — 吞吐: 30.6 tps, 延迟: 0.00s
  97. Moonshot Kimi K3 — 吞吐: 29.4 tps, 延迟: 8362.00s
  98. MiniMax abab5.5-chat — 吞吐: 29.1 tps, 延迟: 6618.00s
  99. Alibaba qwen-mt-plus — 吞吐: 27.9 tps, 延迟: 0.00s
  100. Alibaba qwen3-vl-plus — 吞吐: 25.9 tps, 延迟: 0.00s