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. Meta Muse Spark 1.1 — 吞吐: 42666.7 tps, 延迟: 12200.00s
  2. xAI Grok 4.20 Multi-Agent — 吞吐: 16162.7 tps, 延迟: 7042.00s
  3. OpenAI o3-mini — 吞吐: 13652.2 tps, 延迟: 2031.00s
  4. ByteDance Seedance 2.0 — 吞吐: 3178.7 tps, 延迟: 0.00s
  5. ByteDance Seedance 2.0 Fast — 吞吐: 3114.8 tps, 延迟: 0.00s
  6. ByteDance Seedance 2.5 — 吞吐: 2723.6 tps, 延迟: 0.00s
  7. Meta Muse Spark 1.3 — 吞吐: 2493.3 tps, 延迟: 7958.00s
  8. DeepSeek DeepSeek-V4-Flash 0731 — 吞吐: 1750.0 tps, 延迟: 8659.00s
  9. ByteDance Seedance 2.0 Mini — 吞吐: 1092.5 tps, 延迟: 238079.00s
  10. OpenAI gpt-5-pro — 吞吐: 1000.0 tps, 延迟: 7617.00s
  11. OpenAI GPT-6 Sol — 吞吐: 890.6 tps, 延迟: 1443.33s
  12. xAI grok-4-1-fast-non-reasoning — 吞吐: 666.7 tps, 延迟: 518.00s
  13. OpenAI gpt-3.5-turbo — 吞吐: 426.8 tps, 延迟: 764.75s
  14. Z.ai GLM 5.2 — 吞吐: 426.6 tps, 延迟: 1693.00s
  15. ByteDance Doubao-Seedream-5.0-lite — 吞吐: 377.0 tps, 延迟: 38200.00s
  16. Google Gemini 3.5 Flash — 吞吐: 355.2 tps, 延迟: 6106.40s
  17. Meta Muse Spark 1.2 — 吞吐: 346.7 tps, 延迟: 14614.00s
  18. Google Gemini 3.6 Flash — 吞吐: 339.4 tps, 延迟: 2371.00s
  19. Z.ai GLM 5.3 Flash — 吞吐: 333.0 tps, 延迟: 8095.00s
  20. Google Gemini 3 Flash Preview — 吞吐: 263.5 tps, 延迟: 2083.38s
  21. OpenAI GPT-5 — 吞吐: 260.8 tps, 延迟: 24135.88s
  22. Google Gemini 3.1 Flash-Lite Preview — 吞吐: 246.8 tps, 延迟: 0.00s
  23. Google Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image 🍌) — 吞吐: 237.6 tps, 延迟: 6118.38s
  24. Google Gemini 3.1 Pro Preview — 吞吐: 222.7 tps, 延迟: 11724.70s
  25. Google Gemini 3.8 Flash — 吞吐: 215.5 tps, 延迟: 5552.80s
  26. OpenAI gpt-4-gizmo-* — 吞吐: 207.7 tps, 延迟: 7577.67s
  27. OpenAI gpt-5-nano — 吞吐: 181.2 tps, 延迟: 1493.90s
  28. DeepSeek DeepSeek V4.1 Flash — 吞吐: 178.9 tps, 延迟: 11863.43s
  29. Anthropic Claude Claude Haiku 5.5 — 吞吐: 174.6 tps, 延迟: 2457.24s
  30. OpenAI gpt-5-mini — 吞吐: 173.8 tps, 延迟: 5373.33s
  31. Google Nano Banana (Gemini 2.5 Flash Image 🍌) — 吞吐: 170.6 tps, 延迟: 0.00s
  32. OpenAI GPT-5.6 Luna — 吞吐: 170.5 tps, 延迟: 1248.09s
  33. OpenAI GPT-5.4 Nano — 吞吐: 161.0 tps, 延迟: 1501.17s
  34. OpenAI chatgpt-image-latest — 吞吐: 152.3 tps, 延迟: 43817.00s
  35. Google Gemini 3.7 Flash — 吞吐: 150.9 tps, 延迟: 4132.37s
  36. OpenAI GPT Image 1.5 — 吞吐: 142.1 tps, 延迟: 29167.57s
  37. OpenAI GPT-5.4 Mini — 吞吐: 139.8 tps, 延迟: 12435.68s
  38. Google Gemini Nano Banana 2.1 🍌 — 吞吐: 139.4 tps, 延迟: 0.00s
  39. Google Gemini 4 Argon — 吞吐: 133.5 tps, 延迟: 9787.00s
  40. Z.ai GLM 5 — 吞吐: 116.5 tps, 延迟: 10237.00s
  41. xAI Grok 4.20 — 吞吐: 108.5 tps, 延迟: 0.00s
  42. Anthropic Claude Claude Sonnet 5.5 — 吞吐: 107.1 tps, 延迟: 5712.93s
  43. xAI grok-3 — 吞吐: 106.9 tps, 延迟: 0.00s
  44. OpenAI GPT-5.5 — 吞吐: 103.9 tps, 延迟: 6438.09s
  45. OpenAI GPT-5.6 Sol — 吞吐: 103.1 tps, 延迟: 3035.00s
  46. Anthropic Claude Claude Opus 5.5 — 吞吐: 102.6 tps, 延迟: 3686.61s
  47. xAI Grok 4.3 — 吞吐: 98.8 tps, 延迟: 2527.11s
  48. Meta Muse Wonder 1.0 — 吞吐: 92.4 tps, 延迟: 50723.65s
  49. xAI grok-4-0709 — 吞吐: 92.3 tps, 延迟: 2967.00s
  50. Xiaomi MiMo-V2.6-Flash — 吞吐: 92.3 tps, 延迟: 15946.00s
  51. xAI grok-4-1-fast-reasoning — 吞吐: 89.9 tps, 延迟: 0.00s
  52. OpenAI GPT-5.2 — 吞吐: 87.4 tps, 延迟: 1012.59s
  53. MiniMax minimax-m2 — 吞吐: 87.4 tps, 延迟: 1899.00s
  54. OpenAI gpt-oss-120b — 吞吐: 86.4 tps, 延迟: 7326.00s
  55. xAI grok-4-fast-reasoning — 吞吐: 86.3 tps, 延迟: 0.00s
  56. xAI grok-3-mini — 吞吐: 86.1 tps, 延迟: 0.00s
  57. OpenAI gpt-4o-all — 吞吐: 83.0 tps, 延迟: 7071.50s
  58. OpenAI gpt-image-1-mini — 吞吐: 81.0 tps, 延迟: 13044.25s
  59. Anthropic Claude Claude Sonnet 5 — 吞吐: 79.3 tps, 延迟: 5127.85s
  60. DeepSeek DeepSeek-V4-Pro-0813 — 吞吐: 77.3 tps, 延迟: 2566.00s
  61. Anthropic Claude Claude Opus 5 — 吞吐: 75.2 tps, 延迟: 3891.14s
  62. OpenAI GPT Image 2 Develop — 吞吐: 73.7 tps, 延迟: 36173.80s
  63. xAI Grok Build 0.1 — 吞吐: 72.5 tps, 延迟: 0.00s
  64. OpenAI gpt-4.1-mini — 吞吐: 71.8 tps, 延迟: 2404.00s
  65. Google gemini-2.5-pro — 吞吐: 71.8 tps, 延迟: 6636.00s
  66. OpenAI GPT Image 2.5 Dev — 吞吐: 65.8 tps, 延迟: 93988.00s
  67. OpenAI gpt-4-turbo — 吞吐: 64.7 tps, 延迟: 0.00s
  68. ByteDance Doubao-Seed-2.0-mini 260428 — 吞吐: 64.2 tps, 延迟: 0.00s
  69. OpenAI GPT Image 1.0 — 吞吐: 63.4 tps, 延迟: 20970.12s
  70. Meta Muse Spark 1.2 Contributor — 吞吐: 61.8 tps, 延迟: 1029.00s
  71. Thinking Machines Thinking Machines: Inkling — 吞吐: 61.7 tps, 延迟: 0.00s
  72. xAI Grok 4.5 — 吞吐: 61.2 tps, 延迟: 1344.00s
  73. Google Gemini 3.5 Flash-Lite — 吞吐: 58.4 tps, 延迟: 2122.83s
  74. Google Gemini 3.1 Flash-Lite — 吞吐: 57.9 tps, 延迟: 0.00s
  75. xAI Grok 4.6 — 吞吐: 57.2 tps, 延迟: 2507.17s
  76. OpenAI GPT-6.1 Sol — 吞吐: 57.0 tps, 延迟: 959.81s
  77. MiniMax MiniMax M2.5 — 吞吐: 56.1 tps, 延迟: 11687.00s
  78. OpenAI GPT-5.6 Terra — 吞吐: 54.3 tps, 延迟: 2206.50s
  79. OpenAI gpt-4o-mini — 吞吐: 52.9 tps, 延迟: 1507.50s
  80. Meta Muse Spark 1.3 Contributor — 吞吐: 52.4 tps, 延迟: 1247.00s
  81. ByteDance Doubao-Seed-2.0-mini — 吞吐: 50.4 tps, 延迟: 0.00s
  82. Google Gemini 3 Pro Preview — 吞吐: 50.2 tps, 延迟: 0.00s
  83. Z.ai GLM 4.5 Air — 吞吐: 49.8 tps, 延迟: 1124.00s
  84. Z.ai GLM 5.3 — 吞吐: 47.6 tps, 延迟: 10509.50s
  85. OpenAI GPT-6 Luna — 吞吐: 47.4 tps, 延迟: 1585.47s
  86. Anthropic Claude Claude Opus 4.8 — 吞吐: 46.5 tps, 延迟: 3734.43s
  87. xAI grok-4-fast-non-reasoning — 吞吐: 46.4 tps, 延迟: 526.33s
  88. Moonshot kimi-k2-thinking — 吞吐: 44.5 tps, 延迟: 11202.00s
  89. DeepSeek DeepSeek-V3.2 — 吞吐: 44.4 tps, 延迟: 1910.00s
  90. ByteDance Doubao-Seed-2.1-pro — 吞吐: 41.5 tps, 延迟: 0.00s
  91. OpenAI GPT-5.1 — 吞吐: 40.1 tps, 延迟: 3494.33s
  92. Alibaba Qwen3.8-Max — 吞吐: 40.0 tps, 延迟: 0.00s
  93. OpenAI GPT Image 2.5 Sunburst — 吞吐: 38.7 tps, 延迟: 23305.65s
  94. ByteDance Doubao-Seed-2.1-turbo — 吞吐: 38.5 tps, 延迟: 0.00s
  95. Alibaba Qwen3.5 Plus 2026-02-15 — 吞吐: 35.1 tps, 延迟: 0.00s
  96. Anthropic Claude Claude Opus 4.7 — 吞吐: 35.1 tps, 延迟: 1970.12s
  97. OpenAI gpt-4o-transcribe — 吞吐: 34.9 tps, 延迟: 0.00s
  98. Google gemini-2.5-flash — 吞吐: 34.1 tps, 延迟: 0.00s
  99. xAI Grok 4.7 — 吞吐: 30.2 tps, 延迟: 0.00s
  100. MiniMax abab5.5-chat — 吞吐: 27.3 tps, 延迟: 0.00s