AI 模型速度排行榜
按实测输出吞吐(tokens/秒)与平均延迟排名。
- MiniMax MiniMax H3 — 吞吐: 1013.5 tps, 延迟: 770766.00s
- Google gemini-2.5-pro — 吞吐: 721.9 tps, 延迟: 24393.00s
- OpenAI gpt-5-nano — 吞吐: 654.6 tps, 延迟: 15022.24s
- Google Gemini 3.1 Flash-Lite — 吞吐: 341.8 tps, 延迟: 1387.18s
- xAI Grok 4.20 Multi-Agent — 吞吐: 326.0 tps, 延迟: 0.00s
- Google Gemini Omni Flash Preview — 吞吐: 306.6 tps, 延迟: 99229.00s
- Google Gemini 3.6 Flash — 吞吐: 296.8 tps, 延迟: 5997.27s
- OpenAI GPT-5 — 吞吐: 263.0 tps, 延迟: 18101.91s
- Google Gemini 3.5 Flash — 吞吐: 215.4 tps, 延迟: 7233.34s
- Google Gemini 3 Flash Preview — 吞吐: 182.2 tps, 延迟: 5950.54s
- Google Gemini 3.1 Flash-Lite Preview — 吞吐: 179.5 tps, 延迟: 801.80s
- OpenAI o3-mini — 吞吐: 174.8 tps, 延迟: 4997.00s
- OpenAI GPT-5.4 Mini — 吞吐: 162.1 tps, 延迟: 1206.38s
- Google gemini-2.5-flash — 吞吐: 161.2 tps, 延迟: 28991.95s
- OpenAI GPT Image 1.0 — 吞吐: 153.3 tps, 延迟: 34143.31s
- Google Gemini 3.1 Pro Preview — 吞吐: 152.6 tps, 延迟: 17943.26s
- OpenAI dall-e-3 — 吞吐: 149.2 tps, 延迟: 27980.00s
- OpenAI GPT-5.6 Terra — 吞吐: 142.1 tps, 延迟: 6338.00s
- Google Gemini 3.1 Pro Preview Customtools — 吞吐: 124.8 tps, 延迟: 39148.25s
- OpenAI GPT-5.1 — 吞吐: 124.2 tps, 延迟: 1645.39s
- OpenAI gpt-4-gizmo-* — 吞吐: 123.9 tps, 延迟: 12297.88s
- Anthropic Claude Haiku 4.5 — 吞吐: 123.4 tps, 延迟: 1673.02s
- OpenAI gpt-oss-120b — 吞吐: 119.6 tps, 延迟: 0.00s
- OpenAI gpt-4o-all — 吞吐: 115.1 tps, 延迟: 2348.24s
- OpenAI GPT-5.5 — 吞吐: 110.7 tps, 延迟: 4632.25s
- OpenAI GPT Image 2 Develop — 吞吐: 107.6 tps, 延迟: 37968.59s
- xAI grok-3 — 吞吐: 105.7 tps, 延迟: 0.00s
- OpenAI gpt-5.1-codex — 吞吐: 101.7 tps, 延迟: 1361.00s
- xAI grok-4-1-fast-reasoning — 吞吐: 100.3 tps, 延迟: 0.00s
- xAI grok-3-mini — 吞吐: 96.3 tps, 延迟: 0.00s
- xAI Grok 4.20 (Non-Reasoning) — 吞吐: 94.7 tps, 延迟: 1459.67s
- OpenAI gpt-5-mini — 吞吐: 94.6 tps, 延迟: 4654.09s
- Anthropic Claude Sonnet 5 — 吞吐: 93.7 tps, 延迟: 4818.44s
- OpenAI GPT Image 1.5 — 吞吐: 93.1 tps, 延迟: 31909.50s
- Google Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image 🍌) — 吞吐: 91.1 tps, 延迟: 15552.38s
- OpenAI gpt-4.1 — 吞吐: 90.9 tps, 延迟: 740.71s
- OpenAI gpt-4o-2024-11-20 — 吞吐: 90.3 tps, 延迟: 0.00s
- xAI grok-4-0709 — 吞吐: 88.9 tps, 延迟: 890.00s
- xAI Grok 4.3 — 吞吐: 88.2 tps, 延迟: 2255.21s
- OpenAI o4-mini — 吞吐: 88.0 tps, 延迟: 10853.00s
- Z.ai GLM 5.2 — 吞吐: 87.4 tps, 延迟: 1666.80s
- OpenAI gpt-4o-search-preview — 吞吐: 86.8 tps, 延迟: 0.00s
- MiniMax MiniMax M2.5 — 吞吐: 85.9 tps, 延迟: 0.00s
- OpenAI gpt-5-chat-latest — 吞吐: 80.6 tps, 延迟: 10813.27s
- OpenAI GPT-5.6 Luna — 吞吐: 80.1 tps, 延迟: 3593.70s
- OpenAI GPT-5.4 — 吞吐: 78.8 tps, 延迟: 3151.26s
- Anthropic Claude Opus 4.7 — 吞吐: 75.5 tps, 延迟: 4221.58s
- Anthropic Claude Fable 5 — 吞吐: 74.8 tps, 延迟: 6971.13s
- Google Nano Banana (Gemini 2.5 Flash Image 🍌) — 吞吐: 72.8 tps, 延迟: 7099.50s
- Anthropic Claude Opus 5 — 吞吐: 71.2 tps, 延迟: 13071.91s
- OpenAI gpt-4o-image — 吞吐: 71.0 tps, 延迟: 36383.74s
- DeepSeek DeepSeek-V4-Flash 0731 — 吞吐: 69.5 tps, 延迟: 8642.68s
- Google Gemini 3.7 Flash — 吞吐: 68.9 tps, 延迟: 0.00s
- Alibaba qwen-plus — 吞吐: 68.1 tps, 延迟: 2070.56s
- xAI grok-4-1-fast-non-reasoning — 吞吐: 66.5 tps, 延迟: 773.67s
- xAI grok-4-fast-reasoning — 吞吐: 65.9 tps, 延迟: 0.00s
- xAI Grok 4.20 — 吞吐: 64.2 tps, 延迟: 536.00s
- xAI Grok Build 0.1 — 吞吐: 62.4 tps, 延迟: 397.00s
- Anthropic Claude Sonnet 4.6 — 吞吐: 61.9 tps, 延迟: 3378.92s
- DeepSeek DeepSeek-V4-Pro — 吞吐: 61.5 tps, 延迟: 1390.11s
- Anthropic Claude Opus 4.8 — 吞吐: 61.2 tps, 延迟: 8774.78s
- OpenAI gpt-image-1-mini — 吞吐: 61.0 tps, 延迟: 17320.00s
- OpenAI gpt-5.1-codex-mini — 吞吐: 60.1 tps, 延迟: 0.00s
- OpenAI gpt-4.1-mini — 吞吐: 59.0 tps, 延迟: 3694.03s
- OpenAI GPT-5.6 Sol — 吞吐: 58.9 tps, 延迟: 31500.24s
- Anthropic Claude Sonnet 4.5 — 吞吐: 58.4 tps, 延迟: 2047.05s
- DeepSeek deepseek-v3.1 — 吞吐: 56.9 tps, 延迟: 3098.00s
- OpenAI GPT-5.2 — 吞吐: 56.9 tps, 延迟: 2301.53s
- Google Gemini 3 Pro Preview — 吞吐: 56.9 tps, 延迟: 0.00s
- xAI Grok 4.5 — 吞吐: 54.9 tps, 延迟: 2946.32s
- OpenAI gpt-4o-2024-08-06 — 吞吐: 54.7 tps, 延迟: 0.00s
- OpenAI gpt-4o-mini-transcribe — 吞吐: 50.6 tps, 延迟: 0.00s
- OpenAI gpt-4.1-nano — 吞吐: 49.8 tps, 延迟: 664.07s
- DeepSeek DeepSeek-V3.2 — 吞吐: 49.3 tps, 延迟: 959.20s
- Anthropic Claude Opus 4.6 — 吞吐: 49.3 tps, 延迟: 8110.79s
- xAI Grok 4.6 — 吞吐: 47.2 tps, 延迟: 2666.58s
- OpenAI GPT-5.4 Nano — 吞吐: 46.3 tps, 延迟: 1178.33s
- OpenAI gpt-3.5-turbo-16k — 吞吐: 42.5 tps, 延迟: 0.00s
- Google Gemini 3.5 Flash-Lite — 吞吐: 42.2 tps, 延迟: 1582.56s
- DeepSeek DeepSeek-V4-Pro-0813 — 吞吐: 42.0 tps, 延迟: 254.67s
- Anthropic Claude Opus 4.5 — 吞吐: 41.0 tps, 延迟: 2208.50s
- Google Nano Banana 2 (Gemini 3.1 Flash Image 🍌) — 吞吐: 37.8 tps, 延迟: 44087.46s
- OpenAI GPT Image 2 — 吞吐: 34.9 tps, 延迟: 55690.28s
- Moonshot Kimi K3 — 吞吐: 33.4 tps, 延迟: 5536.00s
- OpenAI GPT-5.5 Pro — 吞吐: 33.0 tps, 延迟: 101140.20s
- ByteDance Doubao-Seed-2.0-pro — 吞吐: 31.7 tps, 延迟: 0.00s
- OpenAI o1 — 吞吐: 30.6 tps, 延迟: 0.00s
- OpenAI gpt-4o — 吞吐: 29.2 tps, 延迟: 4694.00s
- Alibaba qwen-max — 吞吐: 27.0 tps, 延迟: 657.00s
- OpenAI gpt-3.5-turbo — 吞吐: 25.5 tps, 延迟: 920.00s
- OpenAI gpt-4o-mini — 吞吐: 24.2 tps, 延迟: 3378.96s
- MiniMax MiniMax M3 — 吞吐: 23.4 tps, 延迟: 0.00s
- Moonshot Kimi K2.6 — 吞吐: 22.3 tps, 延迟: 929.18s
- Alibaba qwen3-vl-flash — 吞吐: 20.0 tps, 延迟: 0.00s
- Xiaomi MiMo-V2.5 — 吞吐: 19.7 tps, 延迟: 2852.20s
- Google Nano Banana Pro (Gemini 3 Pro Image 🍌) — 吞吐: 17.4 tps, 延迟: 43986.76s
- Yi yi-lightning — 吞吐: 17.1 tps, 延迟: 0.00s
- OpenAI gpt-3.5-turbo-1106 — 吞吐: 17.0 tps, 延迟: 0.00s
- OpenAI gpt-4o-transcribe — 吞吐: 13.4 tps, 延迟: 0.00s
- xAI grok-4-fast-non-reasoning — 吞吐: 10.4 tps, 延迟: 0.00s