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. OpenAI gpt-4-1106-preview — 吞吐: 9000.0 tps, 延迟: 1315.00s
  2. ByteDance Seedance 2.0 Mini — 吞吐: 1432.3 tps, 延迟: 181600.00s
  3. OpenAI gpt-4.1 — 吞吐: 1000.0 tps, 延迟: 1450.00s
  4. Google gemini-2.5-flash — 吞吐: 718.2 tps, 延迟: 5008.33s
  5. TypeSafe Jev 1.13 — 吞吐: 476.3 tps, 延迟: 0.00s
  6. Xiaomi MiMo-V2.6-Pro UltraSpeed — 吞吐: 469.4 tps, 延迟: 4103.00s
  7. ByteDance Seedance 2.0 Fast — 吞吐: 390.4 tps, 延迟: 129721.00s
  8. xAI Grok 4.20 Multi-Agent — 吞吐: 373.3 tps, 延迟: 0.00s
  9. Google Gemini 3.1 Pro Preview — 吞吐: 358.8 tps, 延迟: 14641.86s
  10. Google Gemini 3.6 Flash — 吞吐: 338.9 tps, 延迟: 11955.29s
  11. Google Gemini 3.5 Flash — 吞吐: 327.7 tps, 延迟: 6909.60s
  12. Google Gemini 3 Flash Preview — 吞吐: 261.4 tps, 延迟: 2915.00s
  13. Google Gemini 3.1 Flash-Lite Preview — 吞吐: 255.4 tps, 延迟: 0.00s
  14. OpenAI GPT-5 — 吞吐: 246.5 tps, 延迟: 14919.50s
  15. OpenAI GPT-5.4 Mini — 吞吐: 220.7 tps, 延迟: 7354.00s
  16. Google Gemini 4 Argon — 吞吐: 211.9 tps, 延迟: 21951.83s
  17. Anthropic Claude Claude Haiku 5.5 — 吞吐: 209.5 tps, 延迟: 1106.39s
  18. OpenAI dall-e-3 — 吞吐: 199.3 tps, 延迟: 20953.00s
  19. OpenAI GPT Image 2.5 Dev — 吞吐: 193.5 tps, 延迟: 32299.00s
  20. Google Gemini 3.8 Flash — 吞吐: 183.8 tps, 延迟: 4548.95s
  21. OpenAI gpt-4-gizmo-* — 吞吐: 178.1 tps, 延迟: 3001.50s
  22. Google Gemini 3.1 Pro Preview Customtools — 吞吐: 166.2 tps, 延迟: 0.00s
  23. OpenAI gpt-4o-all — 吞吐: 162.8 tps, 延迟: 5105.12s
  24. ByteDance Doubao-Seedream-5.0-pro — 吞吐: 150.4 tps, 延迟: 77040.67s
  25. OpenAI gpt-audio-mini — 吞吐: 150.0 tps, 延迟: 384.09s
  26. OpenAI chatgpt-image-latest — 吞吐: 146.6 tps, 延迟: 49308.00s
  27. Anthropic Claude Claude Sonnet 5.5 — 吞吐: 143.0 tps, 延迟: 8071.25s
  28. DeepSeek DeepSeek V4.1 Flash — 吞吐: 136.6 tps, 延迟: 4691.07s
  29. OpenAI GPT-5.6 Terra — 吞吐: 132.9 tps, 延迟: 634.00s
  30. Google Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image 🍌) — 吞吐: 130.3 tps, 延迟: 11383.40s
  31. xAI Grok 4.20 — 吞吐: 119.7 tps, 延迟: 0.00s
  32. OpenAI gpt-5-nano — 吞吐: 117.6 tps, 延迟: 1374.67s
  33. OpenAI gpt-4.1-nano — 吞吐: 116.1 tps, 延迟: 611.63s
  34. Alibaba qwen3-coder-flash — 吞吐: 115.8 tps, 延迟: 1616.00s
  35. Meta Muse Wonder 1.0 — 吞吐: 113.7 tps, 延迟: 2232.50s
  36. Google Nano Banana (Gemini 2.5 Flash Image 🍌) — 吞吐: 112.5 tps, 延迟: 12854.00s
  37. ByteDance Doubao-Seed-2.0-mini 260428 — 吞吐: 110.4 tps, 延迟: 0.00s
  38. xAI grok-3 — 吞吐: 104.7 tps, 延迟: 0.00s
  39. OpenAI GPT-5.6 Sol — 吞吐: 104.1 tps, 延迟: 1510.31s
  40. MiniMax minimax-m2.1 — 吞吐: 103.3 tps, 延迟: 2603.00s
  41. Alibaba Qwen3.7 Max — 吞吐: 102.8 tps, 延迟: 3830.00s
  42. OpenAI gpt-5-mini — 吞吐: 99.3 tps, 延迟: 6993.56s
  43. Anthropic Claude Claude Opus 5.5 — 吞吐: 99.1 tps, 延迟: 2913.55s
  44. xAI Grok Build 0.1 — 吞吐: 98.9 tps, 延迟: 0.00s
  45. DeepSeek DeepSeek-V4-Flash 0731 — 吞吐: 95.7 tps, 延迟: 1419.00s
  46. MiniMax MiniMax M2.5 — 吞吐: 95.7 tps, 延迟: 6362.00s
  47. Moonshot Kimi K2.6 — 吞吐: 92.4 tps, 延迟: 9948.00s
  48. DeepSeek DeepSeek-V4-Flash — 吞吐: 92.2 tps, 延迟: 4560.80s
  49. Z.ai GLM 5 — 吞吐: 90.1 tps, 延迟: 6745.25s
  50. OpenAI GPT-5.4 Nano — 吞吐: 89.9 tps, 延迟: 1664.00s
  51. Google gemini-2.5-pro — 吞吐: 89.6 tps, 延迟: 1597.50s
  52. OpenAI GPT Image 2 Develop — 吞吐: 89.5 tps, 延迟: 24253.14s
  53. DeepSeek DeepSeek-V4-Pro-0813 — 吞吐: 89.3 tps, 延迟: 4107.17s
  54. OpenAI GPT-5.5 — 吞吐: 89.1 tps, 延迟: 2767.40s
  55. OpenAI GPT-6.1 Sol — 吞吐: 85.0 tps, 延迟: 1502.24s
  56. Z.ai GLM 5 Turbo — 吞吐: 84.7 tps, 延迟: 12867.00s
  57. Google Gemini 3.7 Flash — 吞吐: 81.9 tps, 延迟: 28409.67s
  58. OpenAI gpt-image-1-mini — 吞吐: 81.8 tps, 延迟: 12904.00s
  59. xAI Grok 4.3 — 吞吐: 81.0 tps, 延迟: 1885.67s
  60. Anthropic Claude Claude Opus 5 — 吞吐: 79.4 tps, 延迟: 1963.39s
  61. OpenAI gpt-3.5-turbo — 吞吐: 79.2 tps, 延迟: 0.00s
  62. Alibaba Qwen3.8-Flash — 吞吐: 78.8 tps, 延迟: 29325.50s
  63. Z.ai GLM 5.2 — 吞吐: 78.5 tps, 延迟: 11716.08s
  64. OpenAI gpt-4o-transcribe-diarize — 吞吐: 71.8 tps, 延迟: 0.00s
  65. OpenAI GPT-6 Luna — 吞吐: 71.7 tps, 延迟: 2561.29s
  66. Anthropic Claude Claude Fable 5.1 — 吞吐: 71.4 tps, 延迟: 3091.00s
  67. xAI Grok 4.7 — 吞吐: 71.1 tps, 延迟: 1756.25s
  68. OpenAI gpt-4.1-mini — 吞吐: 66.6 tps, 延迟: 0.00s
  69. xAI grok-3-mini — 吞吐: 64.1 tps, 延迟: 0.00s
  70. OpenAI gpt-3.5-turbo-16k — 吞吐: 63.9 tps, 延迟: 5737.00s
  71. Google Gemini 3.5 Flash-Lite — 吞吐: 62.4 tps, 延迟: 2167.58s
  72. Anthropic Claude Claude Opus 4.7 — 吞吐: 62.3 tps, 延迟: 1501.42s
  73. xAI grok-4-1-fast-reasoning — 吞吐: 61.7 tps, 延迟: 0.00s
  74. Z.ai GLM 4.5 Air — 吞吐: 61.7 tps, 延迟: 0.00s
  75. xAI grok-4-fast-reasoning — 吞吐: 60.4 tps, 延迟: 0.00s
  76. Meta Muse Image 1.0 — 吞吐: 59.8 tps, 延迟: 25217.00s
  77. Alibaba Qwen3.5 Plus 2026-02-15 — 吞吐: 58.8 tps, 延迟: 1520.00s
  78. OpenAI GPT-5.4 — 吞吐: 57.9 tps, 延迟: 2271.29s
  79. xAI Grok 4.5 — 吞吐: 57.6 tps, 延迟: 1634.50s
  80. OpenAI GPT-5.6 Luna — 吞吐: 53.7 tps, 延迟: 3063.06s
  81. MiniMax MiniMax M2.5 Lightning — 吞吐: 53.3 tps, 延迟: 0.00s
  82. ByteDance Doubao-Seed-2.0-mini — 吞吐: 52.4 tps, 延迟: 0.00s
  83. xAI Grok 4.6 — 吞吐: 52.3 tps, 延迟: 1566.50s
  84. OpenAI gpt-4o-mini-transcribe — 吞吐: 52.3 tps, 延迟: 0.00s
  85. Alibaba Qwen3.6 Plus — 吞吐: 51.0 tps, 延迟: 0.00s
  86. Anthropic Claude Claude Sonnet 5 — 吞吐: 49.7 tps, 延迟: 1572.33s
  87. Google Gemini Nano Banana 2.1 🍌 — 吞吐: 49.4 tps, 延迟: 69423.00s
  88. Alibaba Qwen3.8-Max — 吞吐: 47.8 tps, 延迟: 0.00s
  89. OpenAI GPT Image 1.5 — 吞吐: 46.8 tps, 延迟: 13196.00s
  90. xAI grok-4-0709 — 吞吐: 46.6 tps, 延迟: 2451.57s
  91. Anthropic Claude Claude Opus 4.8 — 吞吐: 46.5 tps, 延迟: 889.00s
  92. OpenAI gpt-oss-120b — 吞吐: 46.2 tps, 延迟: 299489.29s
  93. ByteDance Doubao-Seed-2.1-pro — 吞吐: 45.1 tps, 延迟: 0.00s
  94. xAI grok-4-fast-non-reasoning — 吞吐: 44.4 tps, 延迟: 618.00s
  95. Google Gemini 3.1 Flash-Lite — 吞吐: 44.4 tps, 延迟: 0.00s
  96. OpenAI GPT Image 2.5 Sunburst — 吞吐: 43.3 tps, 延迟: 45296.20s
  97. Anthropic Claude Claude Opus 4.5 — 吞吐: 42.8 tps, 延迟: 3220.17s
  98. xAI grok-4-1-fast-non-reasoning — 吞吐: 42.8 tps, 延迟: 3048.67s
  99. Z.ai GLM 5.1 — 吞吐: 42.7 tps, 延迟: 0.00s
  100. Google Gemini 3 Pro Preview — 吞吐: 42.5 tps, 延迟: 0.00s