AI Model Speed Rankings
Models ranked by measured output throughput (tokens per second) and average latency.
- Moonshot Kimi K2.5 — Throughput: 916.7 tps, Latency: 983.50s
- OpenAI gpt-5-nano — Throughput: 624.9 tps, Latency: 24240.08s
- xAI Grok 4.20 Multi-Agent — Throughput: 358.8 tps, Latency: 0.00s
- Google Gemini 3.1 Flash-Lite — Throughput: 343.8 tps, Latency: 1060.45s
- Google gemini-2.5-pro — Throughput: 275.7 tps, Latency: 26226.97s
- Google Nano Banana 2 Lite (Gemini 3.1 Flash Lite Image 🍌) — Throughput: 260.0 tps, Latency: 5625.92s
- Google Nano Banana (Gemini 2.5 Flash Image 🍌) — Throughput: 247.2 tps, Latency: 8021.00s
- Google Gemini 3.1 Pro Preview — Throughput: 237.9 tps, Latency: 29535.64s
- OpenAI GPT-5.6 Luna — Throughput: 222.2 tps, Latency: 3591.84s
- OpenAI GPT-5 — Throughput: 211.3 tps, Latency: 22619.39s
- Google Gemini 3.5 Flash — Throughput: 200.3 tps, Latency: 7725.31s
- Google Gemini 3 Flash Preview — Throughput: 199.6 tps, Latency: 7603.31s
- Google gemini-2.5-flash — Throughput: 178.0 tps, Latency: 17261.88s
- xAI Grok 4.20 — Throughput: 148.0 tps, Latency: 696.00s
- xAI Grok 4.20 (Non-Reasoning) — Throughput: 145.6 tps, Latency: 0.00s
- OpenAI gpt-4o-all — Throughput: 137.1 tps, Latency: 4958.33s
- xAI Grok 4.5 — Throughput: 126.2 tps, Latency: 5538.48s
- OpenAI GPT-5.6 Terra — Throughput: 121.5 tps, Latency: 4571.65s
- xAI grok-3-mini — Throughput: 113.8 tps, Latency: 0.00s
- Google Gemini 3.1 Pro Preview Customtools — Throughput: 113.5 tps, Latency: 0.00s
- OpenAI GPT-5.3 Codex — Throughput: 113.3 tps, Latency: 4485.00s
- xAI grok-4-1-fast-reasoning — Throughput: 112.2 tps, Latency: 0.00s
- DeepSeek deepseek-chat — Throughput: 110.8 tps, Latency: 0.00s
- xAI Grok 4.3 — Throughput: 106.0 tps, Latency: 1363.63s
- xAI grok-4-0709 — Throughput: 104.2 tps, Latency: 1038.75s
- OpenAI dall-e-3 — Throughput: 103.5 tps, Latency: 60377.00s
- OpenAI GPT Image 1.5 — Throughput: 103.2 tps, Latency: 44204.00s
- xAI grok-4-1-fast-non-reasoning — Throughput: 101.6 tps, Latency: 788.64s
- xAI Grok Build 0.1 — Throughput: 99.8 tps, Latency: 2197.00s
- OpenAI Chat Latest (GPT-5.5 Instant) — Throughput: 98.8 tps, Latency: 0.00s
- Alibaba qwen-plus — Throughput: 92.3 tps, Latency: 1138.71s
- Moonshot Kimi K2.6 — Throughput: 85.2 tps, Latency: 773.50s
- OpenAI GPT Image 1.0 — Throughput: 84.9 tps, Latency: 48977.00s
- OpenAI gpt-4-all — Throughput: 84.4 tps, Latency: 5470.56s
- OpenAI gpt-4-gizmo-* — Throughput: 80.5 tps, Latency: 4098.12s
- Z.ai GLM 5.1 — Throughput: 78.4 tps, Latency: 524.50s
- OpenAI gpt-4.1-mini — Throughput: 76.5 tps, Latency: 903.86s
- OpenAI gpt-5.1-codex-mini — Throughput: 76.4 tps, Latency: 760.50s
- OpenAI GPT‑5.3 Instant — Throughput: 74.9 tps, Latency: 1262.00s
- OpenAI o4-mini — Throughput: 74.6 tps, Latency: 5126.50s
- ByteDance Doubao-Seed-2.0-mini — Throughput: 73.9 tps, Latency: 0.00s
- OpenAI gpt-4o-image — Throughput: 72.7 tps, Latency: 36870.00s
- OpenAI GPT-5.4 — Throughput: 72.6 tps, Latency: 4842.77s
- Anthropic Claude Sonnet 5 — Throughput: 72.2 tps, Latency: 2365.22s
- OpenAI gpt-5.2-chat-latest — Throughput: 72.0 tps, Latency: 3630.38s
- Google Nano Banana 2 (Gemini 3.1 Flash Image 🍌) — Throughput: 71.7 tps, Latency: 16479.81s
- Anthropic Claude Opus 4.7 — Throughput: 68.1 tps, Latency: 2966.60s
- Anthropic Claude Opus 4.5 — Throughput: 67.7 tps, Latency: 3284.22s
- Stepfun step-3.5-flash — Throughput: 67.5 tps, Latency: 194.00s
- OpenAI GPT-5.4 Nano — Throughput: 67.5 tps, Latency: 12255.50s
- Anthropic Claude Fable 5 — Throughput: 67.3 tps, Latency: 5153.04s
- OpenAI GPT-5.5 — Throughput: 67.2 tps, Latency: 3173.92s
- OpenAI GPT-5.4 Mini — Throughput: 67.1 tps, Latency: 8377.00s
- Anthropic Claude Opus 4.8 — Throughput: 66.1 tps, Latency: 2699.80s
- xAI grok-4-fast-reasoning — Throughput: 65.9 tps, Latency: 6836.00s
- Google gemini-2.5-flash-lite — Throughput: 63.7 tps, Latency: 0.00s
- Xiaomi MiMo-V2.5 — Throughput: 60.7 tps, Latency: 4833.59s
- Anthropic Claude Sonnet 4.6 — Throughput: 60.3 tps, Latency: 2587.47s
- OpenAI GPT-5.1 Chat — Throughput: 54.3 tps, Latency: 1174.50s
- OpenAI GPT-5.2 — Throughput: 53.7 tps, Latency: 3459.88s
- Google Gemini 3.1 Flash-Lite Preview — Throughput: 53.4 tps, Latency: 369.00s
- OpenAI GPT-5.1 — Throughput: 51.4 tps, Latency: 1241.08s
- Anthropic Claude Sonnet 4.5 — Throughput: 51.3 tps, Latency: 2450.88s
- Moonshot moonshot-v1-8k — Throughput: 50.4 tps, Latency: 1680.12s
- Anthropic Claude Opus 4.6 — Throughput: 49.2 tps, Latency: 3510.84s
- DeepSeek deepseek-reasoner — Throughput: 47.3 tps, Latency: 314.84s
- Z.ai GLM 5.2 — Throughput: 46.0 tps, Latency: 837.50s
- DeepSeek DeepSeek-V3.2 — Throughput: 41.2 tps, Latency: 1136.00s
- OpenAI gpt-4.1 — Throughput: 40.3 tps, Latency: 2448.67s
- OpenAI gpt-5-mini — Throughput: 39.5 tps, Latency: 10089.19s
- OpenAI gpt-4.1-nano — Throughput: 38.4 tps, Latency: 1158.70s
- Anthropic Claude Opus 4.1 — Throughput: 38.0 tps, Latency: 2566.41s
- OpenAI o1 — Throughput: 37.3 tps, Latency: 0.00s
- OpenAI gpt-3.5-turbo-0125 — Throughput: 37.1 tps, Latency: 15273.50s
- DeepSeek DeepSeek-V4-Flash — Throughput: 33.2 tps, Latency: 1967.17s
- ByteDance Doubao-Seed-2.0-pro — Throughput: 33.0 tps, Latency: 0.00s
- Google Gemini 3 Pro Preview — Throughput: 32.6 tps, Latency: 0.00s
- MiniMax MiniMax-M2.7 — Throughput: 32.3 tps, Latency: 13818.20s
- OpenAI gpt-4o-mini — Throughput: 32.1 tps, Latency: 5427.30s
- OpenAI GPT-5.6 Sol — Throughput: 32.1 tps, Latency: 19184.99s
- Alibaba qwen-max — Throughput: 27.7 tps, Latency: 0.00s
- OpenAI GPT Image 2 — Throughput: 26.5 tps, Latency: 93902.40s
- OpenAI gpt-oss-120b — Throughput: 23.4 tps, Latency: 0.00s
- DeepSeek deepseek-v3 — Throughput: 22.5 tps, Latency: 0.00s
- MiniMax MiniMax M3 — Throughput: 22.1 tps, Latency: 0.00s
- OpenAI GPT Image 2 Develop — Throughput: 22.0 tps, Latency: 83237.63s
- Alibaba qwen3-vl-flash — Throughput: 21.7 tps, Latency: 0.00s
- Moonshot kimi-k2-0711-preview — Throughput: 21.5 tps, Latency: 1090.00s
- OpenAI gpt-4o — Throughput: 21.3 tps, Latency: 6536.20s
- OpenAI GPT-5.4 Pro — Throughput: 19.6 tps, Latency: 1184.12s
- MiniMax MiniMax M2.5 Lightning — Throughput: 18.5 tps, Latency: 795.00s
- Google Nano Banana Pro (Gemini 3 Pro Image 🍌) — Throughput: 18.5 tps, Latency: 27435.99s
- OpenAI o4-mini-deep-research — Throughput: 17.8 tps, Latency: 495.00s
- Moonshot kimi-k2-0905-preview — Throughput: 17.3 tps, Latency: 903.25s
- OpenAI gpt-3.5-turbo-16k — Throughput: 17.2 tps, Latency: 0.00s
- OpenAI gpt-3.5-turbo-1106 — Throughput: 17.2 tps, Latency: 594.00s
- DeepSeek DeepSeek-V4-Pro — Throughput: 17.0 tps, Latency: 5996.67s
- OpenAI gpt-4 — Throughput: 15.1 tps, Latency: 6322.00s
- OpenAI o3 — Throughput: 12.9 tps, Latency: 0.00s
- Moonshot kimi-k2-thinking-turbo — Throughput: 12.7 tps, Latency: 908.50s