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. Moonshot Kimi K2.5 — 吞吐: 660.0 tps, 延迟: 1444.00s
  2. OpenAI gpt-5-nano — 吞吐: 440.8 tps, 延迟: 18354.27s
  3. Google Gemini 3.5 Flash — 吞吐: 392.0 tps, 延迟: 8593.76s
  4. xAI Grok 4.20 Multi-Agent — 吞吐: 358.8 tps, 延迟: 0.00s
  5. Google Gemini 3.1 Flash-Lite — 吞吐: 292.3 tps, 延迟: 1129.52s
  6. Google Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image 🍌) — 吞吐: 266.0 tps, 延迟: 5624.75s
  7. Google gemini-2.5-pro — 吞吐: 243.7 tps, 延迟: 52948.99s
  8. OpenAI GPT-5.6 Luna — 吞吐: 212.9 tps, 延迟: 3720.12s
  9. OpenAI GPT-5 — 吞吐: 212.5 tps, 延迟: 23789.72s
  10. Google Nano Banana (Gemini 2.5 Flash Image 🍌) — 吞吐: 209.7 tps, 延迟: 0.00s
  11. Google Gemini 3 Flash Preview — 吞吐: 205.1 tps, 延迟: 8708.67s
  12. xAI Grok 4.20 (Non-Reasoning) — 吞吐: 202.7 tps, 延迟: 851.58s
  13. xAI Grok 4.20 — 吞吐: 193.5 tps, 延迟: 537.29s
  14. Google Gemini 3.1 Pro Preview — 吞吐: 192.0 tps, 延迟: 43886.36s
  15. Google gemini-2.5-flash — 吞吐: 147.4 tps, 延迟: 14107.14s
  16. OpenAI GPT-5.4 Mini — 吞吐: 140.9 tps, 延迟: 1459.45s
  17. DeepSeek deepseek-chat — 吞吐: 116.7 tps, 延迟: 0.00s
  18. Google Gemini 3.1 Pro Preview Customtools — 吞吐: 113.5 tps, 延迟: 0.00s
  19. OpenAI GPT-5.3 Codex — 吞吐: 113.3 tps, 延迟: 4485.00s
  20. xAI Grok 4.5 — 吞吐: 111.2 tps, 延迟: 3498.61s
  21. xAI grok-4-1-fast-reasoning — 吞吐: 106.6 tps, 延迟: 0.00s
  22. OpenAI dall-e-3 — 吞吐: 103.5 tps, 延迟: 60377.00s
  23. OpenAI GPT Image 1.5 — 吞吐: 103.2 tps, 延迟: 44204.00s
  24. OpenAI gpt-4-gizmo-* — 吞吐: 103.0 tps, 延迟: 5074.47s
  25. OpenAI o4-mini — 吞吐: 102.4 tps, 延迟: 7955.80s
  26. xAI grok-3-mini — 吞吐: 102.1 tps, 延迟: 397.00s
  27. xAI grok-4-0709 — 吞吐: 100.6 tps, 延迟: 1047.43s
  28. xAI Grok Build 0.1 — 吞吐: 99.8 tps, 延迟: 2197.00s
  29. OpenAI gpt-4o-all — 吞吐: 98.9 tps, 延迟: 5900.00s
  30. xAI grok-3 — 吞吐: 98.3 tps, 延迟: 0.00s
  31. xAI Grok 4.3 — 吞吐: 95.6 tps, 延迟: 951.73s
  32. xAI grok-4-1-fast-non-reasoning — 吞吐: 90.2 tps, 延迟: 759.42s
  33. OpenAI GPT Image 1.0 — 吞吐: 89.0 tps, 延迟: 30225.00s
  34. OpenAI GPT-5.4 — 吞吐: 85.5 tps, 延迟: 2005.53s
  35. Alibaba qwen-plus — 吞吐: 84.4 tps, 延迟: 1050.20s
  36. OpenAI gpt-5.2-chat-latest — 吞吐: 80.8 tps, 延迟: 4796.90s
  37. Z.ai GLM 5.1 — 吞吐: 78.3 tps, 延迟: 523.00s
  38. OpenAI GPT-5.6 Terra — 吞吐: 77.0 tps, 延迟: 5467.15s
  39. OpenAI gpt-5.1-codex-mini — 吞吐: 76.4 tps, 延迟: 760.50s
  40. Xiaomi MiMo-V2.5 — 吞吐: 76.1 tps, 延迟: 8047.14s
  41. Anthropic Claude Opus 4.7 — 吞吐: 75.6 tps, 延迟: 2514.81s
  42. xAI grok-4-fast-reasoning — 吞吐: 74.2 tps, 延迟: 0.00s
  43. ByteDance Doubao-Seed-2.0-mini — 吞吐: 73.9 tps, 延迟: 0.00s
  44. OpenAI GPT-5.1 Chat — 吞吐: 71.1 tps, 延迟: 2530.00s
  45. Anthropic Claude Sonnet 5 — 吞吐: 70.9 tps, 延迟: 7203.14s
  46. OpenAI GPT‑5.3 Instant — 吞吐: 69.3 tps, 延迟: 969.91s
  47. DeepSeek DeepSeek-V4-Pro — 吞吐: 68.7 tps, 延迟: 1954.69s
  48. Z.ai GLM 5.2 — 吞吐: 68.7 tps, 延迟: 1024.00s
  49. Stepfun step-3.5-flash — 吞吐: 67.5 tps, 延迟: 194.00s
  50. Anthropic Claude Opus 4.8 — 吞吐: 66.4 tps, 延迟: 3139.75s
  51. OpenAI gpt-4.1-nano — 吞吐: 63.1 tps, 延迟: 1327.94s
  52. Anthropic Claude Fable 5 — 吞吐: 62.7 tps, 延迟: 5324.04s
  53. OpenAI gpt-5-mini — 吞吐: 61.1 tps, 延迟: 14605.82s
  54. Moonshot Kimi K2.6 — 吞吐: 61.1 tps, 延迟: 615.00s
  55. Alibaba qwen3-max — 吞吐: 60.9 tps, 延迟: 1274.75s
  56. OpenAI gpt-4.1 — 吞吐: 60.5 tps, 延迟: 5261.71s
  57. OpenAI gpt-4o-2024-08-06 — 吞吐: 58.8 tps, 延迟: 0.00s
  58. Anthropic Claude Sonnet 4.6 — 吞吐: 58.8 tps, 延迟: 2259.91s
  59. Google Nano Banana 2 (Gemini 3.1 Flash Image 🍌) — 吞吐: 58.3 tps, 延迟: 16999.21s
  60. Anthropic Claude Opus 4.5 — 吞吐: 58.1 tps, 延迟: 2764.47s
  61. OpenAI GPT-5.5 — 吞吐: 57.5 tps, 延迟: 4155.27s
  62. OpenAI gpt-4.1-mini — 吞吐: 57.4 tps, 延迟: 1733.70s
  63. OpenAI GPT-5.4 Nano — 吞吐: 56.0 tps, 延迟: 1970.00s
  64. Moonshot moonshot-v1-8k — 吞吐: 52.6 tps, 延迟: 2196.48s
  65. Anthropic Claude Opus 4.6 — 吞吐: 52.1 tps, 延迟: 2969.15s
  66. OpenAI chatgpt-image-latest — 吞吐: 51.9 tps, 延迟: 48170.40s
  67. OpenAI o4-mini-deep-research — 吞吐: 51.6 tps, 延迟: 697.00s
  68. Google Gemini 3.1 Flash-Lite Preview — 吞吐: 51.6 tps, 延迟: 1031.00s
  69. Anthropic Claude Sonnet 4.5 — 吞吐: 50.5 tps, 延迟: 2136.83s
  70. OpenAI gpt-3.5-turbo — 吞吐: 42.7 tps, 延迟: 1417.20s
  71. OpenAI gpt-4-all — 吞吐: 42.4 tps, 延迟: 5979.83s
  72. OpenAI GPT-5.1 — 吞吐: 41.2 tps, 延迟: 1317.27s
  73. DeepSeek deepseek-reasoner — 吞吐: 41.2 tps, 延迟: 323.46s
  74. OpenAI GPT-5.2 — 吞吐: 39.3 tps, 延迟: 3219.02s
  75. Anthropic Claude Opus 4.1 — 吞吐: 34.3 tps, 延迟: 3012.72s
  76. ByteDance Doubao-Seed-2.0-pro — 吞吐: 33.0 tps, 延迟: 0.00s
  77. MiniMax MiniMax-M2.7 — 吞吐: 32.3 tps, 延迟: 13818.20s
  78. OpenAI GPT-5.6 Sol — 吞吐: 31.1 tps, 延迟: 19161.83s
  79. OpenAI GPT Image 2 — 吞吐: 29.8 tps, 延迟: 70744.19s
  80. Anthropic Claude Haiku 4.5 — 吞吐: 28.4 tps, 延迟: 2254.10s
  81. Alibaba qwen-max — 吞吐: 27.7 tps, 延迟: 0.00s
  82. OpenAI gpt-4o-mini — 吞吐: 25.4 tps, 延迟: 2921.59s
  83. OpenAI gpt-oss-120b — 吞吐: 23.4 tps, 延迟: 0.00s
  84. Alibaba qwen3-vl-flash — 吞吐: 22.5 tps, 延迟: 0.00s
  85. OpenAI GPT Image 2 Develop — 吞吐: 21.8 tps, 延迟: 74675.78s
  86. Z.ai GLM 3 Turbo — 吞吐: 20.8 tps, 延迟: 1055.60s
  87. OpenAI gpt-4o-image — 吞吐: 20.2 tps, 延迟: 174971.97s
  88. OpenAI GPT-5.4 Pro — 吞吐: 18.8 tps, 延迟: 2162.47s
  89. Google Nano Banana Pro (Gemini 3 Pro Image 🍌) — 吞吐: 18.8 tps, 延迟: 33725.90s
  90. MiniMax MiniMax M2.5 Lightning — 吞吐: 18.5 tps, 延迟: 795.00s
  91. DeepSeek DeepSeek-V4-Flash — 吞吐: 17.7 tps, 延迟: 425.80s
  92. OpenAI gpt-3.5-turbo-16k — 吞吐: 17.2 tps, 延迟: 0.00s
  93. OpenAI gpt-3.5-turbo-1106 — 吞吐: 17.2 tps, 延迟: 594.00s
  94. Google gemini-2.5-flash-lite — 吞吐: 16.4 tps, 延迟: 13184.00s
  95. Google Gemini 3 Pro Preview — 吞吐: 15.8 tps, 延迟: 0.00s
  96. Moonshot kimi-k2-0711-preview — 吞吐: 15.2 tps, 延迟: 1436.00s
  97. Moonshot kimi-k2-turbo-preview — 吞吐: 15.2 tps, 延迟: 1148.33s
  98. OpenAI gpt-4o — 吞吐: 15.1 tps, 延迟: 3339.50s
  99. Moonshot kimi-k2-thinking — 吞吐: 14.0 tps, 延迟: 2082.00s
  100. Xiaomi MiMo-V2.5 Pro — 吞吐: 12.9 tps, 延迟: 0.00s