llm.ing

Median output speed

independentrun by Artificial Analysisunit: tokens_per_sDisplay with attribution; no redistribution59 modelssource ↗
AA Intelligence IndexAA Coding IndexAA Agentic IndexMedian output speedMedian time to first token
#ModelScoreObserved
1Mercury 2
513 tok/s
Nov 19, 2026
2Gemini 3.5 Flash Lite
409 tok/s
Nov 19, 2026
3nemotron-3-nano-30b-a3b
303 tok/s
Nov 18, 2026
4Gemini 2.5 Flash-Lite
280 tok/s
Nov 19, 2026
5Muse Spark 1.2
274 tok/s
Nov 19, 2026
6o3-minihigh-effort
251 tok/s
Nov 19, 2026
7Gemini 2.5 Flashthinking
223 tok/s
Nov 19, 2026
8Gemini 3.5 Flash
207 tok/s
Nov 19, 2026
9Claude Haiku 5.5xhigh-effort
182 tok/s
Nov 19, 2026
10GPT-4.1
182 tok/s
Nov 19, 2026
11nvidia-nemotron-nano-9b-v2
174 tok/s
Nov 19, 2026
12DeepSeek R1
160 tok/s
Nov 19, 2026
13Qwen3-Next 80B-A3B Instruct
151 tok/s
Nov 19, 2026
14Qwen3.5 35B-A3B
145 tok/s
Sep 28, 2026
15Qwen3.5 122B-A10Bnon-reasoning
145 tok/s
Nov 19, 2026
16o3
140 tok/s
Nov 18, 2026
17DeepSeek V3.1thinking
136 tok/s
Nov 18, 2026
18GPT-5.6 Luna
133 tok/s
Nov 19, 2026
19GPT-5.2 Codex
132 tok/s
Nov 19, 2026
20GPT-5.1 Codex
127 tok/s
Nov 19, 2026
21GPT-6 Lunaxhigh-effort
115 tok/s
Nov 19, 2026
22Gemini 2.5 Pro
115 tok/s
Nov 19, 2026
23GPT-5
113 tok/s
Sep 22, 2026
24Mistral Large 4
111 tok/s
Oct 29, 2026
25Claude Sonnet 5.5high-effort
105 tok/s
Nov 19, 2026
26GPT-5 Mini
103 tok/s
Nov 19, 2026
27GPT-5.4non-reasoning
98 tok/s
Sep 22, 2026
28GPT-6.2 Solxhigh-effort
97 tok/s
Nov 19, 2026
29GLM-4.7-Flash
96 tok/s
Nov 19, 2026
30Qwen Turbo
96 tok/s
Oct 17, 2026
31Claude Opus 5.5
96 tok/s
Nov 19, 2026
32GPT-5.1
94 tok/s
Sep 1, 2026
33MiniMax-M3
94 tok/s
Nov 18, 2026
34GPT-5.2
93 tok/s
Sep 21, 2026
35Grok Build 0.1
92 tok/s
Nov 19, 2026
36Qwen3-Coder 480B-A35B Instruct
86 tok/s
Nov 19, 2026
37GPT-5.5
83 tok/s
Nov 19, 2026
38Qwen3.5 27Bnon-reasoning
83 tok/s
Aug 27, 2026
39GLM-4.6V
80 tok/s
Nov 18, 2026
40Claude Sonnet 5
78 tok/s
Nov 19, 2026
41GLM-5
71 tok/s
Nov 19, 2026
42Grok 4.7high-effort
70 tok/s
Nov 19, 2026
43GPT-6.1 Terranon-reasoning
67 tok/s
Nov 19, 2026
44Claude Fable 5.1
65 tok/s
Nov 19, 2026
45MiniMax-M2.7
64 tok/s
Sep 2, 2026
46Qwen3.5 397B-A17Bnon-reasoning
58 tok/s
Nov 19, 2026
47MiniMax-M2.5
58 tok/s
Nov 19, 2026
48GPT-6.1 Solxhigh-effort
57 tok/s
Nov 19, 2026
49Qwen3.6 27Bnon-reasoning
57 tok/s
Nov 19, 2026
50MiniMax-M2.1
55 tok/s
Nov 12, 2026
51GLM-5.3-Flash
55 tok/s
Nov 14, 2026
52Claude Opus 4.5 (latest)thinking
55 tok/s
Nov 19, 2026
53Claude Opus 5xhigh-effort
52 tok/s
Nov 19, 2026
54Qwen3.8 27Bnon-reasoning
50 tok/s
Nov 19, 2026
55GPT-6 Astraxhigh-effort
47 tok/s
Nov 19, 2026
56Claude Opus 4.7non-reasoning
46 tok/s
Nov 19, 2026
57Claude Sonnet 4.6
44 tok/s
Nov 19, 2026
58Claude Opus 4.6
40 tok/s
Nov 2, 2026
59GLM-4.5Vthinking
39 tok/s
Nov 19, 2026

Bars on a 0–513.01 tokens_per_s scale. Scores link to model pages; every number keeps its source on the model page.