AI Model Speed Rankings

Models ranked by measured output throughput (tokens per second) and average latency.

  1. OpenAI gpt-5-pro — Throughput: 1000.0 tps, Latency: 1474.50s
  2. OpenAI gpt-4 — Throughput: 1000.0 tps, Latency: 511.00s
  3. ByteDance Seedance 2.0 — Throughput: 891.1 tps, Latency: 364603.00s
  4. Google gemini-2.5-pro — Throughput: 557.9 tps, Latency: 25077.03s
  5. ByteDance Seedance 2.5 — Throughput: 526.4 tps, Latency: 330005.00s
  6. ByteDance Seedance 2.0 Mini — Throughput: 429.2 tps, Latency: 115936.00s
  7. Google Gemini 3.1 Flash-Lite — Throughput: 415.9 tps, Latency: 717.91s
  8. ByteDance Doubao-Seedream-5.0-pro — Throughput: 411.8 tps, Latency: 53348.33s
  9. xAI Grok 4.20 Multi-Agent — Throughput: 386.7 tps, Latency: 0.00s
  10. Google gemini-2.5-flash — Throughput: 371.4 tps, Latency: 394.00s
  11. ByteDance Seedance 2.0 Fast — Throughput: 365.9 tps, Latency: 138400.00s
  12. TypeSafe Jev 1.13 — Throughput: 323.3 tps, Latency: 0.00s
  13. Google Gemini 3.6 Flash — Throughput: 290.1 tps, Latency: 9323.00s
  14. xAI Grok Build 0.1 — Throughput: 258.5 tps, Latency: 0.00s
  15. TypeSafe Jev 1.14 — Throughput: 244.9 tps, Latency: 0.00s
  16. Google Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image 🍌) — Throughput: 214.9 tps, Latency: 6829.57s
  17. Anthropic Claude Claude Haiku 5.5 — Throughput: 208.2 tps, Latency: 1630.38s
  18. Google Gemini 3 Flash Preview — Throughput: 208.2 tps, Latency: 3286.80s
  19. Google Gemini 3.8 Flash — Throughput: 199.9 tps, Latency: 8772.52s
  20. Meta Muse Spark 1.3 — Throughput: 178.7 tps, Latency: 4749.33s
  21. OpenAI GPT-5 — Throughput: 168.7 tps, Latency: 12956.44s
  22. OpenAI GPT-5.6 Luna — Throughput: 163.9 tps, Latency: 2980.92s
  23. Google Gemini 3.1 Pro Preview Customtools — Throughput: 163.3 tps, Latency: 15981.00s
  24. OpenAI gpt-5-nano — Throughput: 159.9 tps, Latency: 2191.00s
  25. Google Nano Banana (Gemini 2.5 Flash Image 🍌) — Throughput: 159.6 tps, Latency: 7693.67s
  26. OpenAI GPT Image 1.5 — Throughput: 159.5 tps, Latency: 43536.00s
  27. Anthropic Claude Claude Sonnet 5 — Throughput: 145.3 tps, Latency: 2898.97s
  28. OpenAI GPT Image 2.5 Dev — Throughput: 144.3 tps, Latency: 28925.33s
  29. OpenAI GPT Image 2 Develop — Throughput: 134.8 tps, Latency: 41515.26s
  30. OpenAI GPT-5.4 Nano — Throughput: 131.0 tps, Latency: 484.19s
  31. OpenAI GPT-5.5 — Throughput: 130.4 tps, Latency: 4526.40s
  32. OpenAI GPT-5.4 Mini — Throughput: 126.1 tps, Latency: 596.70s
  33. Google Gemini 3.5 Flash — Throughput: 120.5 tps, Latency: 4234.46s
  34. DeepSeek DeepSeek V4.1 Flash — Throughput: 119.0 tps, Latency: 16753.76s
  35. Anthropic Claude Claude Opus 4.7 — Throughput: 118.3 tps, Latency: 1862.22s
  36. OpenAI GPT-5.6 Terra — Throughput: 116.4 tps, Latency: 2508.91s
  37. xAI Grok 4.7 — Throughput: 116.3 tps, Latency: 1415.60s
  38. OpenAI GPT-5.2 — Throughput: 116.2 tps, Latency: 472.00s
  39. Google Gemini 3.1 Pro Preview — Throughput: 114.2 tps, Latency: 62331.00s
  40. Anthropic Claude Claude Opus 5.5 — Throughput: 111.8 tps, Latency: 2947.66s
  41. OpenAI dall-e-3 — Throughput: 111.6 tps, Latency: 35823.00s
  42. DeepSeek DeepSeek-V4-Flash 0731 — Throughput: 108.6 tps, Latency: 1632.00s
  43. OpenAI GPT-6 Luna — Throughput: 105.3 tps, Latency: 5030.93s
  44. Google Gemini Nano Banana 2.1 🍌 — Throughput: 103.2 tps, Latency: 25480.00s
  45. OpenAI gpt-image-1-mini — Throughput: 102.1 tps, Latency: 15352.00s
  46. OpenAI GPT-6 Sol — Throughput: 98.3 tps, Latency: 990.09s
  47. Anthropic Claude Claude Sonnet 5.5 — Throughput: 97.9 tps, Latency: 2450.81s
  48. OpenAI o4-mini — Throughput: 92.4 tps, Latency: 2316.00s
  49. OpenAI gpt-4o-mini-transcribe — Throughput: 91.0 tps, Latency: 0.00s
  50. OpenAI gpt-5-mini — Throughput: 90.9 tps, Latency: 1210.35s
  51. Z.ai GLM 5.3 Flash — Throughput: 89.5 tps, Latency: 1602.50s
  52. Google Gemini 4 Argon — Throughput: 89.0 tps, Latency: 19339.71s
  53. Google Nano Banana 2 (Gemini 3.1 Flash Image 🍌) — Throughput: 88.2 tps, Latency: 12903.11s
  54. xAI grok-3 — Throughput: 85.8 tps, Latency: 0.00s
  55. OpenAI gpt-4o-transcribe-diarize — Throughput: 85.1 tps, Latency: 0.00s
  56. xAI grok-4-1-fast-reasoning — Throughput: 84.1 tps, Latency: 1213.30s
  57. xAI Grok 4.6 — Throughput: 83.5 tps, Latency: 2444.50s
  58. xAI Grok 4.3 — Throughput: 83.1 tps, Latency: 1893.33s
  59. OpenAI gpt-4.1-mini — Throughput: 82.3 tps, Latency: 2009.00s
  60. OpenAI GPT-5.1 — Throughput: 82.0 tps, Latency: 1000.42s
  61. Anthropic Claude Claude Fable 5.1 — Throughput: 81.2 tps, Latency: 4853.00s
  62. MiniMax MiniMax M3 — Throughput: 76.9 tps, Latency: 1735.33s
  63. xAI Grok 4.20 — Throughput: 76.7 tps, Latency: 0.00s
  64. xAI Grok 4.5 — Throughput: 76.1 tps, Latency: 3627.78s
  65. MiniMax MiniMax M2.5 Lightning — Throughput: 76.1 tps, Latency: 0.00s
  66. xAI grok-4-0709 — Throughput: 74.1 tps, Latency: 0.00s
  67. Anthropic Claude Claude Opus 5 — Throughput: 73.9 tps, Latency: 2231.39s
  68. xAI grok-3-mini — Throughput: 73.8 tps, Latency: 0.00s
  69. xAI grok-4-fast-reasoning — Throughput: 73.7 tps, Latency: 0.00s
  70. Alibaba Qwen3.8-Flash — Throughput: 71.7 tps, Latency: 0.00s
  71. xAI grok-4-fast-non-reasoning — Throughput: 71.4 tps, Latency: 533.00s
  72. Anthropic Claude Claude Fable 5 — Throughput: 70.2 tps, Latency: 3336.75s
  73. Google Gemini 3.7 Flash — Throughput: 65.3 tps, Latency: 17182.00s
  74. xAI grok-4-1-fast-non-reasoning — Throughput: 64.8 tps, Latency: 505.58s
  75. OpenAI GPT-6 Astra — Throughput: 64.5 tps, Latency: 1424.68s
  76. OpenAI GPT Image 2.5 Flare — Throughput: 63.4 tps, Latency: 17323.60s
  77. OpenAI GPT Image 2 — Throughput: 61.7 tps, Latency: 32593.33s
  78. OpenAI GPT Image 2.5 Sunburst — Throughput: 60.3 tps, Latency: 32194.21s
  79. OpenAI GPT-5.3 Codex — Throughput: 59.4 tps, Latency: 697.19s
  80. OpenAI gpt-4-turbo-preview — Throughput: 59.0 tps, Latency: 3596.00s
  81. xAI Grok 4.20 (Non-Reasoning) — Throughput: 56.5 tps, Latency: 0.00s
  82. OpenAI GPT-5.6 Sol — Throughput: 55.8 tps, Latency: 4128.09s
  83. DeepSeek DeepSeek-V3.2 — Throughput: 55.3 tps, Latency: 1424.00s
  84. DeepSeek DeepSeek-V4-Pro-0813 — Throughput: 50.2 tps, Latency: 1475.00s
  85. MiniMax abab5.5-chat — Throughput: 49.6 tps, Latency: 0.00s
  86. OpenAI GPT-5.4 — Throughput: 47.6 tps, Latency: 3631.00s
  87. Google Gemini 3.1 Flash-Lite Preview — Throughput: 47.5 tps, Latency: 959.25s
  88. Z.ai GLM 5.3 — Throughput: 46.7 tps, Latency: 0.00s
  89. Moonshot moonshot-v1-32k — Throughput: 45.9 tps, Latency: 0.00s
  90. Z.ai GLM 5 — Throughput: 45.7 tps, Latency: 0.00s
  91. Alibaba qwen-plus-latest — Throughput: 45.3 tps, Latency: 1290.00s
  92. Anthropic Claude Claude Haiku 4.5 — Throughput: 44.6 tps, Latency: 1481.77s
  93. Anthropic Claude Claude Opus 4.8 — Throughput: 43.2 tps, Latency: 1391.17s
  94. Moonshot moonshot-v1-8k — Throughput: 42.6 tps, Latency: 0.00s
  95. Moonshot Kimi K2.6 — Throughput: 42.4 tps, Latency: 3091.00s
  96. OpenAI GPT-5.5 Pro — Throughput: 41.5 tps, Latency: 433.00s
  97. Alibaba qwen-plus-2025-07-14 — Throughput: 40.9 tps, Latency: 0.00s
  98. Google Nano Banana Pro (Gemini 3 Pro Image 🍌) — Throughput: 39.9 tps, Latency: 20201.79s
  99. Moonshot kimi-k2-thinking — Throughput: 38.1 tps, Latency: 950.00s
  100. OpenAI GPT-6.1 Sol — Throughput: 37.0 tps, Latency: 6983.51s