AI Model Speed Rankings

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

  1. Z.ai GLM 5.2 — Throughput: 25000.0 tps, Latency: 889.00s
  2. xAI Grok 4.20 Multi-Agent — Throughput: 5082.4 tps, Latency: 3320.00s
  3. ByteDance Seedance 2.0 Mini — Throughput: 1774.0 tps, Latency: 0.00s
  4. xAI grok-4-fast-non-reasoning — Throughput: 1000.0 tps, Latency: 484.00s
  5. ByteDance Seedance 2.0 Fast — Throughput: 782.0 tps, Latency: 139251.00s
  6. Google Gemini 3.6 Flash — Throughput: 559.1 tps, Latency: 6646.50s
  7. ByteDance Doubao-Seedream-5.0-pro — Throughput: 523.5 tps, Latency: 30994.00s
  8. ByteDance Seedance 2.0 — Throughput: 429.5 tps, Latency: 253535.00s
  9. Google Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image 🍌) — Throughput: 269.3 tps, Latency: 5533.14s
  10. Google Gemini 3 Flash Preview — Throughput: 257.8 tps, Latency: 2628.35s
  11. Google Gemini 3.1 Pro Preview — Throughput: 248.0 tps, Latency: 14766.47s
  12. Google Gemini 3.5 Flash — Throughput: 241.5 tps, Latency: 4051.00s
  13. OpenAI dall-e-2 — Throughput: 224.7 tps, Latency: 18577.00s
  14. Anthropic Claude Claude Haiku 5.5 — Throughput: 219.4 tps, Latency: 1336.67s
  15. Z.ai GLM 5.3 Flash — Throughput: 212.7 tps, Latency: 549.00s
  16. Google Gemini 3.1 Flash-Lite Preview — Throughput: 185.4 tps, Latency: 0.00s
  17. Google Gemini 3.8 Flash — Throughput: 182.0 tps, Latency: 5323.45s
  18. OpenAI GPT-5.2 Pro — Throughput: 176.5 tps, Latency: 3694.00s
  19. DeepSeek DeepSeek V4.1 Flash — Throughput: 167.5 tps, Latency: 1856.07s
  20. xAI Grok Build 0.1 — Throughput: 163.7 tps, Latency: 942.00s
  21. xAI grok-4-1-fast-non-reasoning — Throughput: 156.2 tps, Latency: 504.00s
  22. OpenAI dall-e-3 — Throughput: 145.0 tps, Latency: 28784.00s
  23. OpenAI gpt-audio-mini — Throughput: 143.7 tps, Latency: 384.00s
  24. OpenAI gpt-5-mini — Throughput: 141.7 tps, Latency: 1725.14s
  25. xAI Grok 4.3 — Throughput: 136.5 tps, Latency: 1984.17s
  26. MiniMax minimax-m2.1 — Throughput: 129.9 tps, Latency: 1483.00s
  27. Anthropic Claude Claude Sonnet 5.5 — Throughput: 128.1 tps, Latency: 7904.56s
  28. OpenAI GPT Image 1.5 — Throughput: 127.2 tps, Latency: 35936.00s
  29. Google Gemini Nano Banana 2.1 🍌 — Throughput: 126.0 tps, Latency: 0.00s
  30. OpenAI GPT-6 Sol — Throughput: 123.5 tps, Latency: 707.67s
  31. OpenAI gpt-5-nano — Throughput: 122.8 tps, Latency: 3559.73s
  32. Z.ai GLM 5.3 — Throughput: 120.1 tps, Latency: 0.00s
  33. OpenAI GPT-5.4 Mini — Throughput: 119.9 tps, Latency: 12252.47s
  34. ByteDance Doubao-Seed-2.0-mini 260428 — Throughput: 109.2 tps, Latency: 0.00s
  35. xAI grok-4-fast-reasoning — Throughput: 109.1 tps, Latency: 0.00s
  36. Google Nano Banana (Gemini 2.5 Flash Image 🍌) — Throughput: 108.3 tps, Latency: 10918.00s
  37. xAI grok-3 — Throughput: 106.5 tps, Latency: 0.00s
  38. xAI Grok 4.20 — Throughput: 104.1 tps, Latency: 0.00s
  39. Google Gemini 3.7 Flash — Throughput: 98.8 tps, Latency: 4393.73s
  40. Meta Muse Image 1.0 — Throughput: 98.5 tps, Latency: 16214.04s
  41. MiniMax MiniMax M3 — Throughput: 98.1 tps, Latency: 3011.50s
  42. Thinking Machines Thinking Machines: Inkling — Throughput: 88.0 tps, Latency: 0.00s
  43. OpenAI GPT Image 1.0 — Throughput: 87.7 tps, Latency: 71185.00s
  44. OpenAI o3-mini — Throughput: 86.3 tps, Latency: 0.00s
  45. DeepSeek DeepSeek-V4-Pro-0813 — Throughput: 83.8 tps, Latency: 2058.83s
  46. OpenAI GPT-5.4 — Throughput: 82.3 tps, Latency: 1268.10s
  47. OpenAI GPT-5.2 — Throughput: 81.2 tps, Latency: 1071.50s
  48. OpenAI GPT-5.5 — Throughput: 80.3 tps, Latency: 2438.43s
  49. xAI Grok 4.20 (Non-Reasoning) — Throughput: 75.8 tps, Latency: 0.00s
  50. Anthropic Claude Claude Sonnet 5 — Throughput: 75.3 tps, Latency: 1984.95s
  51. xAI grok-3-mini — Throughput: 73.3 tps, Latency: 0.00s
  52. Alibaba Qwen3.7 Flash — Throughput: 69.6 tps, Latency: 631.00s
  53. Meta Muse Spark 1.3 — Throughput: 68.7 tps, Latency: 0.00s
  54. MiniMax MiniMax-M2.7 — Throughput: 68.2 tps, Latency: 1108.00s
  55. xAI Grok 4.7 — Throughput: 67.3 tps, Latency: 897.50s
  56. OpenAI GPT-5.4 Nano — Throughput: 66.8 tps, Latency: 1951.67s
  57. DeepSeek DeepSeek-V4-Flash 0731 — Throughput: 65.7 tps, Latency: 433.00s
  58. OpenAI gpt-3.5-turbo — Throughput: 65.6 tps, Latency: 0.00s
  59. Google Gemini 3.1 Flash-Lite — Throughput: 63.8 tps, Latency: 0.00s
  60. xAI Grok 4.6 — Throughput: 62.7 tps, Latency: 7014.53s
  61. MiniMax MiniMax-M2.7 Highspeed — Throughput: 61.1 tps, Latency: 1666.00s
  62. xAI grok-4-0709 — Throughput: 61.0 tps, Latency: 1142.00s
  63. Anthropic Claude Claude Opus 5 — Throughput: 60.5 tps, Latency: 3379.62s
  64. OpenAI GPT Image 2.5 Sunburst — Throughput: 60.4 tps, Latency: 35757.15s
  65. Alibaba Qwen3.8-Flash — Throughput: 59.9 tps, Latency: 0.00s
  66. MiniMax MiniMax M2.5 — Throughput: 59.8 tps, Latency: 1193.00s
  67. Alibaba Qwen3.6 Plus — Throughput: 59.8 tps, Latency: 1795.00s
  68. OpenAI GPT-6.1 Sol — Throughput: 59.6 tps, Latency: 909.92s
  69. OpenAI gpt-image-1-mini — Throughput: 58.9 tps, Latency: 16029.29s
  70. Alibaba Qwen3.5 Plus 2026-02-15 — Throughput: 58.8 tps, Latency: 1480.00s
  71. Anthropic Claude Claude Sonnet 4.6 — Throughput: 58.6 tps, Latency: 16738.75s
  72. xAI grok-4-1-fast-reasoning — Throughput: 58.5 tps, Latency: 0.00s
  73. OpenAI gpt-4o-transcribe-diarize — Throughput: 58.1 tps, Latency: 0.00s
  74. Alibaba Qwen3.7 Max — Throughput: 55.5 tps, Latency: 903.00s
  75. Google Gemini 4 Argon — Throughput: 54.7 tps, Latency: 12372.00s
  76. Alibaba Qwen3.8-Max — Throughput: 53.1 tps, Latency: 2531.00s
  77. OpenAI GPT-5 — Throughput: 52.0 tps, Latency: 23793.57s
  78. ByteDance Doubao-Seed-2.0-mini — Throughput: 51.5 tps, Latency: 0.00s
  79. Anthropic Claude Claude Opus 5.5 — Throughput: 50.4 tps, Latency: 4299.35s
  80. OpenAI GPT-5.5 Pro — Throughput: 48.4 tps, Latency: 1063.00s
  81. OpenAI GPT-6 Luna — Throughput: 47.8 tps, Latency: 2229.77s
  82. Anthropic Claude Claude Opus 4.8 — Throughput: 45.6 tps, Latency: 2603.00s
  83. Google Nano Banana 2 (Gemini 3.1 Flash Image 🍌) — Throughput: 45.4 tps, Latency: 9599.40s
  84. Anthropic Claude Claude Fable 5.1 — Throughput: 45.3 tps, Latency: 2563.00s
  85. Moonshot kimi-k2-thinking — Throughput: 41.7 tps, Latency: 671.00s
  86. MiniMax minimax-m2 — Throughput: 40.3 tps, Latency: 0.00s
  87. Anthropic Claude Claude Opus 4.7 — Throughput: 38.6 tps, Latency: 2220.68s
  88. OpenAI GPT Image 2.5 Flare — Throughput: 37.1 tps, Latency: 23990.55s
  89. OpenAI gpt-4o-transcribe — Throughput: 36.7 tps, Latency: 0.00s
  90. OpenAI gpt-3.5-turbo-16k — Throughput: 35.9 tps, Latency: 0.00s
  91. Moonshot Kimi K2.6 — Throughput: 35.7 tps, Latency: 14874.09s
  92. MiniMax MiniMax M2.5 Lightning — Throughput: 34.3 tps, Latency: 0.00s
  93. OpenAI GPT Image 2 — Throughput: 33.6 tps, Latency: 23092.60s
  94. Google Gemini 3.5 Flash-Lite — Throughput: 33.4 tps, Latency: 1070.40s
  95. ByteDance Doubao-Seed-2.1-pro — Throughput: 32.9 tps, Latency: 0.00s
  96. Alibaba qwen-plus-2025-07-14 — Throughput: 32.9 tps, Latency: 0.00s
  97. DeepSeek DeepSeek-V3.2 — Throughput: 32.7 tps, Latency: 1989.00s
  98. MiniMax MiniMax-M2.1-lightning — Throughput: 30.6 tps, Latency: 0.00s
  99. Moonshot Kimi K3 — Throughput: 29.4 tps, Latency: 8362.00s
  100. MiniMax abab5.5-chat — Throughput: 29.1 tps, Latency: 6618.00s