AI 模型速度排行榜
按实测输出吞吐(tokens/秒)与平均延迟排名。
- OpenAI gpt-4.1-nano — 吞吐: 2000.0 tps, 延迟: 1879.00s
- ByteDance Doubao-Seedream-4.5 — 吞吐: 1115.2 tps, 延迟: 12912.00s
- ByteDance Seedance 2.0 Fast — 吞吐: 823.4 tps, 延迟: 132255.00s
- ByteDance Seedance 2.0 Mini — 吞吐: 811.3 tps, 延迟: 122167.00s
- OpenAI GPT-5 — 吞吐: 705.6 tps, 延迟: 12247.50s
- ByteDance Doubao-Seedream-4.5 — 吞吐: 480.4 tps, 延迟: 36188.00s
- Alibaba qwen-mt-turbo — 吞吐: 442.4 tps, 延迟: 296.00s
- Google Gemini 3.1 Flash-Lite — 吞吐: 438.7 tps, 延迟: 2874.84s
- xAI Grok 4.20 Multi-Agent — 吞吐: 409.6 tps, 延迟: 0.00s
- Google Gemini 3.5 Flash — 吞吐: 313.3 tps, 延迟: 15224.88s
- ByteDance Doubao-Seedream-5.0-pro — 吞吐: 312.3 tps, 延迟: 50982.17s
- Google Gemini 3.6 Flash — 吞吐: 308.5 tps, 延迟: 4233.39s
- xAI grok-4-fast-non-reasoning — 吞吐: 285.7 tps, 延迟: 711.00s
- ByteDance Doubao-Seedream-5.0-lite — 吞吐: 236.0 tps, 延迟: 61021.00s
- Google Gemini 3.1 Flash-Lite Preview — 吞吐: 229.3 tps, 延迟: 999.71s
- ByteDance Seedance 2.5 — 吞吐: 223.3 tps, 延迟: 487587.00s
- OpenAI gpt-5-nano — 吞吐: 212.9 tps, 延迟: 958.42s
- Google Gemini 3.8 Flash — 吞吐: 193.8 tps, 延迟: 5395.47s
- Google Gemini 3.1 Pro Preview — 吞吐: 190.1 tps, 延迟: 19005.71s
- Google Gemini 3 Flash Preview — 吞吐: 188.9 tps, 延迟: 6712.42s
- OpenAI o3 — 吞吐: 166.2 tps, 延迟: 0.00s
- OpenAI GPT Image 2.5 Dev — 吞吐: 163.1 tps, 延迟: 34294.95s
- OpenAI GPT Image 1.5 — 吞吐: 162.6 tps, 延迟: 39793.00s
- Google gemini-2.5-flash — 吞吐: 151.5 tps, 延迟: 0.00s
- OpenAI dall-e-3 — 吞吐: 142.7 tps, 延迟: 29253.50s
- Anthropic Claude Claude Haiku 4.5 — 吞吐: 140.5 tps, 延迟: 7027.09s
- OpenAI GPT-5.2 — 吞吐: 132.3 tps, 延迟: 24072.56s
- OpenAI gpt-image-1-mini — 吞吐: 131.0 tps, 延迟: 27802.50s
- Anthropic Claude Claude Sonnet 5 — 吞吐: 128.5 tps, 延迟: 8699.69s
- Anthropic Claude Claude Opus 4.8 — 吞吐: 128.4 tps, 延迟: 3486.50s
- Anthropic Claude Claude Fable 5 — 吞吐: 124.8 tps, 延迟: 21304.75s
- Z.ai GLM 5.2 — 吞吐: 117.0 tps, 延迟: 1933.00s
- OpenAI o4-mini — 吞吐: 115.6 tps, 延迟: 5140.00s
- xAI Grok 4.3 — 吞吐: 114.9 tps, 延迟: 946.81s
- OpenAI GPT-5.4 — 吞吐: 113.8 tps, 延迟: 20717.23s
- xAI grok-4-0709 — 吞吐: 112.0 tps, 延迟: 846.62s
- Google Gemini 3.1 Pro Preview Customtools — 吞吐: 109.3 tps, 延迟: 67514.33s
- xAI grok-3-mini — 吞吐: 109.1 tps, 延迟: 0.00s
- xAI grok-3 — 吞吐: 108.6 tps, 延迟: 504.00s
- DeepSeek DeepSeek V4.1 Flash — 吞吐: 105.3 tps, 延迟: 444.50s
- OpenAI gpt-5-mini — 吞吐: 103.8 tps, 延迟: 2677.92s
- xAI grok-4-1-fast-reasoning — 吞吐: 100.9 tps, 延迟: 0.00s
- Z.ai GLM 5.3 Flash — 吞吐: 97.2 tps, 延迟: 423.00s
- OpenAI GPT Image 1.0 — 吞吐: 97.2 tps, 延迟: 49616.33s
- Anthropic Claude Claude Opus 5 — 吞吐: 95.3 tps, 延迟: 48995.35s
- OpenAI GPT Image 2 Develop — 吞吐: 94.2 tps, 延迟: 27015.19s
- DeepSeek DeepSeek-V4-Flash — 吞吐: 87.6 tps, 延迟: 3528.27s
- xAI grok-4-1-fast-non-reasoning — 吞吐: 83.4 tps, 延迟: 1003.33s
- DeepSeek DeepSeek-V4-Flash 0731 — 吞吐: 82.6 tps, 延迟: 1547.26s
- xAI Grok 4.20 — 吞吐: 80.7 tps, 延迟: 0.00s
- xAI grok-4-fast-reasoning — 吞吐: 80.0 tps, 延迟: 0.00s
- Google gemini-2.5-pro — 吞吐: 79.8 tps, 延迟: 50125.00s
- Google Gemini 3 Pro Preview — 吞吐: 79.3 tps, 延迟: 0.00s
- OpenAI GPT-5.4 Mini — 吞吐: 78.5 tps, 延迟: 6476.37s
- xAI Grok 4.5 — 吞吐: 78.2 tps, 延迟: 3062.47s
- xAI Grok Build 0.1 — 吞吐: 78.0 tps, 延迟: 0.00s
- Google Gemini 3.7 Flash — 吞吐: 76.1 tps, 延迟: 11871.59s
- OpenAI GPT-5.6 Terra — 吞吐: 75.2 tps, 延迟: 17090.47s
- DeepSeek DeepSeek-V4-Pro-0813 — 吞吐: 75.1 tps, 延迟: 2125.78s
- OpenAI GPT Image 2.5 Flare — 吞吐: 72.9 tps, 延迟: 22050.49s
- OpenAI Chat Latest (GPT-5.5 Instant) — 吞吐: 72.8 tps, 延迟: 0.00s
- Moonshot kimi-k2-thinking — 吞吐: 72.7 tps, 延迟: 392.00s
- Moonshot Kimi K2.6 — 吞吐: 72.0 tps, 延迟: 919.00s
- Anthropic Claude Claude Opus 4.7 — 吞吐: 71.2 tps, 延迟: 1593.61s
- Z.ai GLM 5.3 — 吞吐: 70.7 tps, 延迟: 0.00s
- Anthropic Claude Claude Sonnet 4.6 — 吞吐: 70.2 tps, 延迟: 7046.03s
- Alibaba qwen3-vl-plus — 吞吐: 69.9 tps, 延迟: 651.00s
- OpenAI GPT-5.1 — 吞吐: 68.7 tps, 延迟: 24058.77s
- Thinking Machines Thinking Machines: Inkling — 吞吐: 68.6 tps, 延迟: 0.00s
- OpenAI gpt-4o-2024-08-06 — 吞吐: 66.0 tps, 延迟: 0.00s
- Google Gemini 3.5 Flash-Lite — 吞吐: 65.8 tps, 延迟: 3749.10s
- OpenAI GPT Image 2.5 Sunburst — 吞吐: 65.1 tps, 延迟: 38125.57s
- Z.ai GLM 5.1 — 吞吐: 62.3 tps, 延迟: 938.00s
- Alibaba qwen3-235b-a22b-thinking-2507 — 吞吐: 61.5 tps, 延迟: 347.00s
- xAI Grok 4.6 — 吞吐: 61.4 tps, 延迟: 3801.38s
- OpenAI GPT-5.6 Luna — 吞吐: 61.3 tps, 延迟: 2665.51s
- OpenAI gpt-4-gizmo-* — 吞吐: 59.3 tps, 延迟: 9681.70s
- Z.ai GLM 5 — 吞吐: 58.6 tps, 延迟: 824.00s
- MiniMax MiniMax M3 — 吞吐: 57.3 tps, 延迟: 1254.00s
- MiniMax MiniMax-M2.7 Highspeed — 吞吐: 54.8 tps, 延迟: 0.00s
- Google Nano Banana (Gemini 2.5 Flash Image 🍌) — 吞吐: 54.5 tps, 延迟: 7943.93s
- OpenAI gpt-3.5-turbo — 吞吐: 53.2 tps, 延迟: 0.00s
- Z.ai GLM 4.7 — 吞吐: 47.1 tps, 延迟: 418.00s
- OpenAI gpt-4o-mini — 吞吐: 46.7 tps, 延迟: 2821.50s
- DeepSeek deepseek-v3.1 — 吞吐: 45.2 tps, 延迟: 2699.00s
- Anthropic Claude Claude Sonnet 4.5 — 吞吐: 44.6 tps, 延迟: 1938.55s
- MiniMax MiniMax-M2.1-lightning — 吞吐: 44.4 tps, 延迟: 0.00s
- OpenAI gpt-4o — 吞吐: 43.2 tps, 延迟: 2472.50s
- Alibaba qwen-plus-latest — 吞吐: 42.3 tps, 延迟: 0.00s
- Alibaba qvq-plus — 吞吐: 40.0 tps, 延迟: 487.00s
- Google Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image 🍌) — 吞吐: 39.6 tps, 延迟: 37234.80s
- OpenAI GPT-5.5 — 吞吐: 38.7 tps, 延迟: 1068.56s
- Anthropic Claude Claude Fable 5.1 — 吞吐: 38.4 tps, 延迟: 7178.00s
- OpenAI GPT Image 2 — 吞吐: 37.3 tps, 延迟: 37501.14s
- ByteDance Doubao-Seed-2.0-Code — 吞吐: 36.2 tps, 延迟: 0.00s
- Alibaba qwen3-vl-flash — 吞吐: 35.3 tps, 延迟: 0.00s
- Moonshot Kimi K3 — 吞吐: 34.5 tps, 延迟: 7762.25s
- DeepSeek deepseek-v3 — 吞吐: 33.7 tps, 延迟: 4064.55s
- Anthropic Claude Claude Opus 4.6 — 吞吐: 33.5 tps, 延迟: 2473.15s
- OpenAI GPT-5.6 Sol — 吞吐: 29.3 tps, 延迟: 4090.29s