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-5-pro — 吞吐: 2000.0 tps, 延迟: 5089.50s
- ByteDance Seedance 2.0 — 吞吐: 1160.2 tps, 延迟: 280047.00s
- TypeSafe Jev — 吞吐: 856.7 tps, 延迟: 0.00s
- ByteDance Seedance 2.0 Mini — 吞吐: 788.3 tps, 延迟: 191655.50s
- ByteDance Seedance 2.5 — 吞吐: 536.0 tps, 延迟: 172947.25s
- OpenAI GPT-5 — 吞吐: 416.8 tps, 延迟: 11470.75s
- Google gemini-2.5-pro — 吞吐: 396.4 tps, 延迟: 3108.67s
- xAI Grok 4.20 Multi-Agent — 吞吐: 388.4 tps, 延迟: 2882.00s
- Google Gemini 3.6 Flash — 吞吐: 334.1 tps, 延迟: 9881.71s
- Google Gemini 3.5 Flash — 吞吐: 321.5 tps, 延迟: 6213.42s
- OpenAI GPT Image 2.5 Dev — 吞吐: 292.3 tps, 延迟: 21381.09s
- Google Gemini 3 Flash Preview — 吞吐: 265.5 tps, 延迟: 2379.36s
- OpenAI chatgpt-image-latest — 吞吐: 231.3 tps, 延迟: 18840.00s
- Google Gemini 3.8 Flash — 吞吐: 213.8 tps, 延迟: 4131.47s
- ByteDance Doubao-Seedream-5.0-pro — 吞吐: 197.1 tps, 延迟: 38308.00s
- DeepSeek DeepSeek-V4-Flash 0731 — 吞吐: 175.3 tps, 延迟: 1282.00s
- OpenAI GPT-5.6 Luna — 吞吐: 162.4 tps, 延迟: 2175.38s
- OpenAI gpt-4-gizmo-* — 吞吐: 159.4 tps, 延迟: 2513.50s
- Google Nano Banana (Gemini 2.5 Flash Image 🍌) — 吞吐: 159.3 tps, 延迟: 0.00s
- OpenAI GPT-6 Sol — 吞吐: 156.6 tps, 延迟: 5388.50s
- OpenAI gpt-oss-120b — 吞吐: 147.3 tps, 延迟: 0.00s
- DeepSeek DeepSeek V4.1 Flash — 吞吐: 142.0 tps, 延迟: 0.00s
- OpenAI gpt-5-nano — 吞吐: 136.9 tps, 延迟: 2128.67s
- xAI grok-3-mini — 吞吐: 136.3 tps, 延迟: 1182.00s
- OpenAI gpt-4-turbo-preview — 吞吐: 130.2 tps, 延迟: 2823.50s
- OpenAI GPT Image 1.5 — 吞吐: 126.9 tps, 延迟: 52290.29s
- xAI Grok Build 0.1 — 吞吐: 123.6 tps, 延迟: 0.00s
- OpenAI GPT Image 1.0 — 吞吐: 119.2 tps, 延迟: 52338.00s
- Anthropic Claude Claude Sonnet 5.5 — 吞吐: 112.8 tps, 延迟: 3515.55s
- xAI Grok 4.3 — 吞吐: 107.8 tps, 延迟: 2379.64s
- xAI grok-3 — 吞吐: 107.2 tps, 延迟: 0.00s
- OpenAI GPT-5.1 — 吞吐: 105.6 tps, 延迟: 2973.71s
- xAI grok-4-1-fast-reasoning — 吞吐: 105.3 tps, 延迟: 0.00s
- OpenAI gpt-5-search-api — 吞吐: 104.8 tps, 延迟: 0.00s
- OpenAI GPT-5.5 — 吞吐: 102.6 tps, 延迟: 2122.00s
- xAI grok-4-0709 — 吞吐: 99.9 tps, 延迟: 0.00s
- Alibaba Qwen3.8-Flash — 吞吐: 92.6 tps, 延迟: 0.00s
- xAI Grok 4.20 — 吞吐: 92.1 tps, 延迟: 0.00s
- Google Gemini 3.1 Flash-Lite — 吞吐: 92.0 tps, 延迟: 3328.67s
- Z.ai GLM 5.1 — 吞吐: 90.7 tps, 延迟: 1165.00s
- OpenAI GPT-5.4 Mini — 吞吐: 90.0 tps, 延迟: 1780.50s
- OpenAI gpt-image-1-mini — 吞吐: 88.9 tps, 延迟: 13369.75s
- xAI Grok 4.5 — 吞吐: 88.0 tps, 延迟: 1810.12s
- DeepSeek DeepSeek-V4-Flash — 吞吐: 87.7 tps, 延迟: 765.00s
- Thinking Machines Thinking Machines: Inkling — 吞吐: 85.6 tps, 延迟: 0.00s
- Alibaba qwen3-8b — 吞吐: 82.2 tps, 延迟: 309.00s
- Z.ai GLM 5.2 — 吞吐: 75.5 tps, 延迟: 524.50s
- Xiaomi MiMo-V2.5 — 吞吐: 74.8 tps, 延迟: 3432.33s
- Moonshot kimi-k2-thinking — 吞吐: 73.3 tps, 延迟: 413.00s
- OpenAI gpt-4o-2024-08-06 — 吞吐: 73.0 tps, 延迟: 0.00s
- DeepSeek DeepSeek-V4-Pro-0813 — 吞吐: 72.8 tps, 延迟: 1198.45s
- Google Gemini 3 Pro Preview — 吞吐: 72.2 tps, 延迟: 0.00s
- OpenAI GPT-5.2 — 吞吐: 70.1 tps, 延迟: 1158.33s
- Z.ai GLM 5 — 吞吐: 69.2 tps, 延迟: 643.50s
- xAI Grok 4.20 (Non-Reasoning) — 吞吐: 68.9 tps, 延迟: 0.00s
- Z.ai GLM 4.7 — 吞吐: 68.1 tps, 延迟: 414.00s
- Moonshot Kimi K3 — 吞吐: 62.2 tps, 延迟: 3359.50s
- Anthropic Claude Claude Opus 5 — 吞吐: 62.1 tps, 延迟: 3283.00s
- xAI Grok 4.7 — 吞吐: 59.5 tps, 延迟: 824.00s
- xAI grok-4-fast-reasoning — 吞吐: 58.8 tps, 延迟: 0.00s
- Google Gemini 3.7 Flash — 吞吐: 58.2 tps, 延迟: 7258.80s
- xAI grok-4-fast-non-reasoning — 吞吐: 54.1 tps, 延迟: 502.00s
- OpenAI GPT-6 Luna — 吞吐: 54.0 tps, 延迟: 5436.25s
- Alibaba qwen3-vl-flash — 吞吐: 53.0 tps, 延迟: 0.00s
- Alibaba qwen3-vl-flash-2025-10-15 — 吞吐: 52.0 tps, 延迟: 0.00s
- DeepSeek deepseek-v3.1 — 吞吐: 51.1 tps, 延迟: 571.00s
- Xiaomi MiMo-V2.6-Pro — 吞吐: 49.9 tps, 延迟: 8175.42s
- xAI Grok 4.6 — 吞吐: 49.7 tps, 延迟: 3215.50s
- OpenAI GPT-6.1 Sol — 吞吐: 46.9 tps, 延迟: 7206.55s
- DeepSeek deepseek-v3 — 吞吐: 46.1 tps, 延迟: 2468.25s
- Google Nano Banana Pro (Gemini 3 Pro Image 🍌) — 吞吐: 45.2 tps, 延迟: 26634.50s
- MiniMax MiniMax M3 — 吞吐: 44.9 tps, 延迟: 2801.00s
- Anthropic Claude Claude Sonnet 5 — 吞吐: 44.8 tps, 延迟: 3341.11s
- Alibaba qvq-plus — 吞吐: 44.4 tps, 延迟: 0.00s
- MiniMax minimax-m2 — 吞吐: 43.8 tps, 延迟: 0.00s
- Anthropic Claude Claude Sonnet 4.5 — 吞吐: 43.3 tps, 延迟: 1469.31s
- Anthropic Claude Claude Fable 5.1 — 吞吐: 42.5 tps, 延迟: 21305.50s
- MiniMax MiniMax M2.5 Lightning — 吞吐: 42.4 tps, 延迟: 1500.00s
- OpenAI GPT-5.4 Nano — 吞吐: 41.7 tps, 延迟: 3628.40s
- Anthropic Claude Claude Opus 5.5 — 吞吐: 41.2 tps, 延迟: 4334.50s
- ByteDance Doubao-Seed-2.1-pro — 吞吐: 40.7 tps, 延迟: 0.00s
- OpenAI GPT Image 2 — 吞吐: 39.5 tps, 延迟: 26077.07s
- OpenAI GPT-5.4 Pro — 吞吐: 36.1 tps, 延迟: 607.00s
- DeepSeek DeepSeek-V3.2 — 吞吐: 32.8 tps, 延迟: 807.50s
- OpenAI gpt-5-mini — 吞吐: 31.9 tps, 延迟: 6058.25s
- Moonshot Kimi K2.5 — 吞吐: 30.1 tps, 延迟: 335.50s
- OpenAI GPT Image 2 Develop — 吞吐: 29.7 tps, 延迟: 47693.67s
- ByteDance Doubao-Seed-2.0-pro — 吞吐: 28.6 tps, 延迟: 0.00s
- Anthropic Claude Claude Opus 4.8 — 吞吐: 28.4 tps, 延迟: 9599.67s
- OpenAI GPT Image 2.5 Sunburst — 吞吐: 28.3 tps, 延迟: 33160.89s
- ByteDance Doubao-Seed-2.0-mini — 吞吐: 27.4 tps, 延迟: 0.00s
- Alibaba qwen-coder-plus — 吞吐: 27.2 tps, 延迟: 0.00s
- Google Gemini 3.5 Flash-Lite — 吞吐: 26.7 tps, 延迟: 880.50s
- OpenAI GPT Image 2.5 Flare — 吞吐: 26.3 tps, 延迟: 19257.44s
- Alibaba qwen-coder-plus-1106 — 吞吐: 25.5 tps, 延迟: 0.00s
- Anthropic Claude Claude Opus 4.5 — 吞吐: 24.5 tps, 延迟: 4819.39s
- Google Gemini 3.1 Pro Preview — 吞吐: 23.5 tps, 延迟: 18867.83s
- MiniMax minimax-m2.1 — 吞吐: 23.4 tps, 延迟: 0.00s
- Anthropic Claude Claude Sonnet 4.6 — 吞吐: 22.7 tps, 延迟: 28610.50s
- Xiaomi MiMo-V2.5 Pro — 吞吐: 21.3 tps, 延迟: 1123.50s