llm.ing

Median output speed

independentrun by Artificial Analysisunit: tokens_per_sDisplay with attribution; no redistribution47 modelssource ↗
AA Intelligence IndexAA Coding IndexAA Agentic IndexMedian output speedMedian time to first token
#ModelScoreObserved
1Mercury 2
702 tok/s
Sep 21, 2026
2Gemini 3.5 Flash Lite
402 tok/s
Sep 21, 2026
3Gemini 2.5 Flash-Lite
328 tok/s
Sep 21, 2026
4Gemini 3.5 Flash
244 tok/s
Sep 21, 2026
5Muse Spark 1.2
234 tok/s
Sep 20, 2026
6Gemini 2.5 Flashthinking
230 tok/s
Sep 21, 2026
7MiniMax-M3
227 tok/s
Sep 21, 2026
8o3-minihigh-effort
224 tok/s
Sep 21, 2026
9GPT-5.1 Codex
196 tok/s
Jul 25, 2026
10GPT-4.1
190 tok/s
Sep 21, 2026
11Qwen3-Next 80B-A3B Instruct
178 tok/s
Sep 21, 2026
12GPT-5.6 Luna
166 tok/s
Sep 21, 2026
13nemotron-3-nano-30b-a3b
160 tok/s
Sep 21, 2026
14nvidia-nemotron-nano-9b-v2
154 tok/s
Sep 21, 2026
15GPT-5.2 Codex
152 tok/s
Jul 25, 2026
16Qwen3.5 35B-A3B
147 tok/s
Sep 21, 2026
17Gemini 2.5 Pro
145 tok/s
Sep 21, 2026
18Qwen3.5 122B-A10Bnon-reasoning
145 tok/s
Sep 21, 2026
19o3
144 tok/s
Sep 21, 2026
20GPT-5.5
115 tok/s
Sep 21, 2026
21GPT-5
113 tok/s
Sep 21, 2026
22GPT-5.4non-reasoning
111 tok/s
Sep 21, 2026
23MiniMax-M2.1
111 tok/s
Sep 21, 2026
24GPT-5.2
101 tok/s
Sep 21, 2026
25GPT-5 Mini
100 tok/s
Sep 21, 2026
26Qwen Turbo
100 tok/s
Sep 21, 2026
27MiniMax-M2.5
100 tok/s
Sep 21, 2026
28GPT-5.1
94 tok/s
Sep 1, 2026
29GLM-5.3-Flash
94 tok/s
Sep 21, 2026
30Qwen3.5 397B-A17Bnon-reasoning
91 tok/s
Sep 21, 2026
31Qwen3.5 27Bnon-reasoning
83 tok/s
Aug 27, 2026
32GLM-4.7-Flash
80 tok/s
Sep 21, 2026
33Claude Sonnet 5
79 tok/s
Sep 21, 2026
34GLM-5
77 tok/s
Sep 21, 2026
35GLM-4.6V
75 tok/s
Sep 21, 2026
36GPT-6 Astra
71 tok/s
Sep 21, 2026
37Claude Fable 5.1
70 tok/s
Sep 21, 2026
38Qwen3.6 27Bnon-reasoning
64 tok/s
Sep 21, 2026
39MiniMax-M2.7
64 tok/s
Sep 2, 2026
40Claude Opus 5xhigh-effort
58 tok/s
Sep 21, 2026
41Grok Build 0.1
52 tok/s
Sep 21, 2026
42Claude Opus 4.5 (latest)thinking
51 tok/s
Sep 21, 2026
43Qwen3-Coder 480B-A35B Instruct
48 tok/s
Sep 21, 2026
44Claude Opus 4.7non-reasoning
45 tok/s
Sep 21, 2026
45Claude Sonnet 4.6
44 tok/s
Sep 21, 2026
46Claude Opus 4.6
42 tok/s
Sep 21, 2026
47GLM-4.5Vthinking
41 tok/s
Sep 20, 2026

Bars on a 0–702.37 tokens_per_s scale. Scores link to model pages; every number keeps its source on the model page.