Median output speed

independentrun by Artificial Analysisunit: tokens_per_sDisplay with attribution; no redistributionobserved Jul 24, 202642 modelssource ↗

#	Model	Score
1	Mercury 2Inception	958 tok/s
2	Gemini 3.5 Flash LiteGoogle	450 tok/s
3	Gemini 2.5 Flash-LiteGoogle	240 tok/s
4	Gemini 2.5 FlashthinkingGoogle	231 tok/s
5	o3-miniOpenAI	220 tok/s
6	GPT-5.1 CodexOpenAI	196 tok/s
7	Qwen3-Next 80B-A3B InstructAlibaba	183 tok/s
8	Gemini 3.5 FlashGoogle	181 tok/s
9	GPT-5.6 LunaOpenAI	173 tok/s
10	GPT-5.2 CodexOpenAI	168 tok/s
11	Qwen3.5 122B-A10Bnon-reasoningAlibaba	145 tok/s
12	Gemini 2.5 ProGoogle	139 tok/s
13	o3OpenAI	136 tok/s
14	Qwen3.5 35B-A3BAlibaba	130 tok/s
15	nvidia-nemotron-nano-9b-v2Nvidia	116 tok/s
16	GPT-4.1OpenAI	115 tok/s
17	Qwen TurboAlibaba	104 tok/s
18	GLM-4.7-FlashZhipu AI	104 tok/s
19	nemotron-3-nano-30b-a3bNvidia	103 tok/s
20	MiniMax-M3MiniMax (minimax.io)	96 tok/s
21	GPT-5OpenAI	95 tok/s
22	Grok Build 0.1xAI	93 tok/s
23	GPT-5 MiniOpenAI	89 tok/s
24	GPT-5.1OpenAI	88 tok/s
25	GPT-5.4non-reasoningOpenAI	88 tok/s
26	Qwen3.5 27Bnon-reasoningAlibaba	83 tok/s
27	MiniMax-M2.1MiniMax (minimax.io)	77 tok/s
28	GPT-5.2OpenAI	76 tok/s
29	Claude Sonnet 5Anthropic	74 tok/s
30	MiniMax-M2.5MiniMax (minimax.io)	73 tok/s
31	GPT-5.5OpenAI	70 tok/s
32	Qwen3.5 397B-A17BAlibaba	64 tok/s
33	GLM-4.5VthinkingZhipu AI	63 tok/s
34	Qwen3-Coder 480B-A35B InstructAlibaba	59 tok/s
35	MiniMax-M2.7MiniMax (minimax.io)	56 tok/s
36	Qwen3.6 27Bnon-reasoningAlibaba	55 tok/s
37	Claude Opus 4.7non-reasoningAnthropic	46 tok/s
38	Claude Opus 4.5 (latest)thinkingAnthropic	46 tok/s
39	GLM-5Zhipu AI	46 tok/s
40	GLM-4.6VZhipu AI	44 tok/s
41	Claude Sonnet 4.6Anthropic	44 tok/s
42	Claude Opus 4.6Anthropic	41 tok/s

Bars on a 0–1000 tokens_per_s scale. Scores link to model pages; every number keeps its source on the model page.