AI Model Speed Rankings

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

  1. OpenAI gpt-3.5-turbo — Throughput: 27000.0 tps, Latency: 1574.00s
  2. Meta Muse Spark 1.3 — Throughput: 2552.0 tps, Latency: 10402.50s
  3. TypeSafe Jev 1.14 — Throughput: 1912.9 tps, Latency: 0.00s
  4. OpenAI gpt-5-pro — Throughput: 1000.0 tps, Latency: 1132.00s
  5. ByteDance Seedance 2.0 — Throughput: 927.1 tps, Latency: 350446.00s
  6. OpenAI GPT Image 2 Develop — Throughput: 876.2 tps, Latency: 37000.00s
  7. ByteDance Seedance 2.0 Mini — Throughput: 693.0 tps, Latency: 157153.00s
  8. ByteDance Seedance 2.5 — Throughput: 654.0 tps, Latency: 166513.00s
  9. ByteDance Seedance 2.0 Fast — Throughput: 646.2 tps, Latency: 168534.00s
  10. OpenAI gpt-5-search-api — Throughput: 500.0 tps, Latency: 1859.00s
  11. Google Gemini 3.6 Flash — Throughput: 435.1 tps, Latency: 4807.20s
  12. xAI Grok 4.20 Multi-Agent — Throughput: 404.7 tps, Latency: 0.00s
  13. ByteDance Doubao-Seedream-5.0-pro — Throughput: 329.4 tps, Latency: 54643.00s
  14. Google Nano Banana (Gemini 2.5 Flash Image 🍌) — Throughput: 279.1 tps, Latency: 0.00s
  15. Google Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image 🍌) — Throughput: 269.4 tps, Latency: 5449.50s
  16. OpenAI gpt-5-nano — Throughput: 266.7 tps, Latency: 1106.93s
  17. Google Gemini 3.5 Flash — Throughput: 258.2 tps, Latency: 11892.39s
  18. OpenAI GPT-5 — Throughput: 242.3 tps, Latency: 3561.00s
  19. OpenAI dall-e-2 — Throughput: 234.4 tps, Latency: 17808.00s
  20. Google Gemini 3 Flash Preview — Throughput: 220.0 tps, Latency: 3141.50s
  21. Google Gemini 3.1 Pro Preview — Throughput: 214.0 tps, Latency: 12107.18s
  22. Google Gemini 3.1 Flash-Lite Preview — Throughput: 207.3 tps, Latency: 1954.60s
  23. Anthropic Claude Claude Haiku 5.5 — Throughput: 199.4 tps, Latency: 3668.46s
  24. OpenAI GPT-5.4 Mini — Throughput: 189.3 tps, Latency: 492.00s
  25. Google Gemini 3.1 Pro Preview Customtools — Throughput: 165.5 tps, Latency: 45484.33s
  26. Google Gemini Nano Banana 2.1 🍌 — Throughput: 161.6 tps, Latency: 0.00s
  27. OpenAI GPT Image 1.5 — Throughput: 158.0 tps, Latency: 18746.00s
  28. OpenAI GPT-5.4 Nano — Throughput: 157.8 tps, Latency: 600.57s
  29. OpenAI chatgpt-image-latest — Throughput: 155.2 tps, Latency: 42453.00s
  30. Meta Muse Wonder 1.0 — Throughput: 150.3 tps, Latency: 0.00s
  31. Google gemini-2.5-flash — Throughput: 147.8 tps, Latency: 9881.28s
  32. MiniMax MiniMax M3 — Throughput: 136.3 tps, Latency: 1221.00s
  33. Anthropic Claude Claude Sonnet 5.5 — Throughput: 134.5 tps, Latency: 1791.00s
  34. xAI Grok Build 0.1 — Throughput: 128.6 tps, Latency: 0.00s
  35. OpenAI gpt-5-mini — Throughput: 122.9 tps, Latency: 1434.22s
  36. DeepSeek DeepSeek-V4-Flash 0731 — Throughput: 120.6 tps, Latency: 45529.00s
  37. OpenAI GPT-5.1 — Throughput: 119.4 tps, Latency: 1019.00s
  38. OpenAI GPT-5.6 Terra — Throughput: 117.3 tps, Latency: 2415.40s
  39. Google Gemini 3.8 Flash — Throughput: 117.3 tps, Latency: 13086.00s
  40. xAI Grok 4.20 — Throughput: 114.0 tps, Latency: 878.00s
  41. OpenAI GPT-5.5 — Throughput: 109.7 tps, Latency: 3616.42s
  42. Anthropic Claude Claude Opus 5.5 — Throughput: 109.3 tps, Latency: 2024.29s
  43. Anthropic Claude Claude Opus 4.7 — Throughput: 106.8 tps, Latency: 2147.56s
  44. Google gemini-2.5-pro — Throughput: 99.5 tps, Latency: 3906.44s
  45. OpenAI gpt-audio — Throughput: 97.2 tps, Latency: 0.00s
  46. OpenAI GPT-6.1 Terra — Throughput: 92.8 tps, Latency: 972.52s
  47. Google Gemini 3.5 Flash-Lite — Throughput: 86.4 tps, Latency: 3843.96s
  48. OpenAI GPT Image 2.5 Dev — Throughput: 85.0 tps, Latency: 46147.00s
  49. ByteDance Doubao-Seed-2.0-mini 260428 — Throughput: 81.8 tps, Latency: 0.00s
  50. OpenAI GPT-5.4 — Throughput: 80.9 tps, Latency: 1932.00s
  51. OpenAI gpt-4o-2024-08-06 — Throughput: 80.3 tps, Latency: 3176.00s
  52. xAI grok-4-fast-reasoning — Throughput: 79.3 tps, Latency: 0.00s
  53. Anthropic Claude Claude Sonnet 5 — Throughput: 78.3 tps, Latency: 1852.26s
  54. Anthropic Claude Claude Fable 5.1 — Throughput: 75.0 tps, Latency: 2873.73s
  55. Alibaba Qwen3.7 Max — Throughput: 74.9 tps, Latency: 1916.00s
  56. Alibaba Qwen3.8-Flash — Throughput: 74.8 tps, Latency: 1836.00s
  57. OpenAI GPT-5.6 Sol — Throughput: 74.0 tps, Latency: 5261.75s
  58. xAI Grok 4.3 — Throughput: 74.0 tps, Latency: 1305.00s
  59. OpenAI GPT-5.6 Luna — Throughput: 73.4 tps, Latency: 1932.28s
  60. OpenAI gpt-4o — Throughput: 73.2 tps, Latency: 745.50s
  61. Anthropic Claude Claude Fable 5 — Throughput: 72.0 tps, Latency: 2232.88s
  62. xAI grok-3-mini — Throughput: 71.5 tps, Latency: 0.00s
  63. DeepSeek DeepSeek V4.1 Flash — Throughput: 69.5 tps, Latency: 33874.79s
  64. xAI grok-4-1-fast-reasoning — Throughput: 69.1 tps, Latency: 0.00s
  65. xAI grok-3 — Throughput: 68.4 tps, Latency: 0.00s
  66. Anthropic Claude Claude Opus 5 — Throughput: 66.9 tps, Latency: 3852.52s
  67. OpenAI GPT-6.1 Sol — Throughput: 66.3 tps, Latency: 2357.00s
  68. OpenAI GPT Image 2.5 Flare — Throughput: 65.3 tps, Latency: 12263.69s
  69. xAI Grok 4.20 (Non-Reasoning) — Throughput: 64.6 tps, Latency: 0.00s
  70. Alibaba Qwen3.5 Plus 2026-02-15 — Throughput: 62.5 tps, Latency: 1652.00s
  71. xAI grok-4-0709 — Throughput: 62.2 tps, Latency: 3442.50s
  72. Anthropic Claude Claude Opus 4.8 — Throughput: 60.3 tps, Latency: 1632.40s
  73. Google Gemini 4 Argon — Throughput: 60.2 tps, Latency: 25322.64s
  74. Google Gemini 3.7 Flash — Throughput: 59.6 tps, Latency: 27198.00s
  75. Alibaba Qwen3.6 Plus — Throughput: 58.9 tps, Latency: 1240.00s
  76. OpenAI gpt-4-turbo — Throughput: 58.7 tps, Latency: 4278.50s
  77. OpenAI gpt-4o-mini-transcribe — Throughput: 58.1 tps, Latency: 0.00s
  78. Moonshot Kimi K3.1 Pro — Throughput: 58.0 tps, Latency: 10794.00s
  79. xAI Grok 4.7 — Throughput: 57.5 tps, Latency: 2573.28s
  80. xAI Grok 4.6 — Throughput: 57.4 tps, Latency: 3484.83s
  81. xAI Grok 4.5 — Throughput: 55.9 tps, Latency: 3811.86s
  82. Alibaba Qwen3.8-Max — Throughput: 54.8 tps, Latency: 0.00s
  83. Google Gemini 3.1 Flash-Lite — Throughput: 54.1 tps, Latency: 944.00s
  84. xAI grok-4-1-fast-non-reasoning — Throughput: 50.6 tps, Latency: 1478.00s
  85. Anthropic Claude Claude Sonnet 4.6 — Throughput: 49.3 tps, Latency: 1167.21s
  86. OpenAI GPT-5.2 — Throughput: 48.3 tps, Latency: 757.25s
  87. OpenAI GPT-6 Luna — Throughput: 47.5 tps, Latency: 1108.86s
  88. Google Nano Banana Pro (Gemini 3 Pro Image 🍌) — Throughput: 46.6 tps, Latency: 21643.44s
  89. OpenAI GPT Image 2.5 Sunburst — Throughput: 46.0 tps, Latency: 33185.09s
  90. Google Gemini 3 Pro Preview — Throughput: 45.6 tps, Latency: 3233.00s
  91. Moonshot Kimi K3 — Throughput: 43.1 tps, Latency: 62386.20s
  92. OpenAI GPT-5.3 Codex — Throughput: 42.9 tps, Latency: 12474.00s
  93. Alibaba qwen-plus-2025-07-14 — Throughput: 40.1 tps, Latency: 0.00s
  94. OpenAI GPT-6 Astra — Throughput: 39.9 tps, Latency: 1071.50s
  95. OpenAI gpt-4o-transcribe — Throughput: 39.4 tps, Latency: 0.00s
  96. ByteDance Doubao-Seed-2.1-turbo — Throughput: 38.7 tps, Latency: 42277.00s
  97. Thinking Machines Thinking Machines: Inkling — Throughput: 38.6 tps, Latency: 1822.00s
  98. Moonshot Kimi K2.6 — Throughput: 38.0 tps, Latency: 12856.00s
  99. DeepSeek DeepSeek-V4-Pro-0813 — Throughput: 37.7 tps, Latency: 2718.00s
  100. Anthropic Claude Claude Haiku 4.5 — Throughput: 37.6 tps, Latency: 1117.67s