TPSTokens Per Second

AI Model Benchmarks

459 scores across 12 benchmarks — quality, coding, reasoning, and speed.

Output throughput in tokens per second measured on standard prompts. Higher = faster responses. Source: artificialanalysis.ai, ~Feb 2026.

Model
tok/s
1
Nemotron-Labs Diffusion 8BNVIDIA
865
2
Gemini 3.1 Flash LiteGoogle DeepMind
360
3
Step-3.5-Flash-BaseStepFun
300
4
Gemini 2.0 FlashGoogle DeepMind
228
5
GPT-5.4 NanoOpenAI
210
6
Gemini 2.5 FlashGoogle DeepMind
198
7
Claude Haiku 4.5Anthropic
187
8
SonarPerplexity AI
175
9
Gemma 4 26B A4B ITGoogle DeepMind
163.09
10
Llama 4 ScoutMeta AI
160
11
Grok 4 minixAI
145
12
Grok 3 minixAI
138
13
GPT-5.4 miniOpenAI
132
14
Sonar ReasoningPerplexity AI
120
15
o4-miniOpenAI
115
16
GPT-4.1OpenAI
112
17
o3-miniOpenAI
108
18
GPT-4oOpenAI
98
19
Sonar ProPerplexity AI
95
20
Qwen3.5-35B-A3BAlibaba / Qwen
92.51
21
Llama 4 MaverickMeta AI
90
22
GPT-5.4OpenAI
85
23
Claude Sonnet 4.5Anthropic
82
24
Claude Sonnet 4.6Anthropic
80
25
Claude 3.5 SonnetAnthropic
78
26
Mistral Medium 3Mistral AI
75
27
Mistral Large 2Mistral AI
72
28
DeepSeek V3DeepSeek
70
29
Claude 3.7 SonnetAnthropic
68
30
Qwen3 72BAlibaba / Qwen
65
31
Sonar Reasoning ProPerplexity AI
65
32
Gemini 2.5 ProGoogle DeepMind
62
33
Phi-4Microsoft
60
34
DeepSeek R1DeepSeek
58
35
Grok 3xAI
55
36
QwQ-32BAlibaba / Qwen
50
37
Gemini 1.5 ProGoogle DeepMind
45
38
GPT-4.5OpenAI
42
39
Claude Opus 4Anthropic
40
40
Claude Opus 4.6Anthropic
38
41
o3OpenAI
32
42
Grok 4xAI
28
43
Llama 3.1 405BMeta AI
26
44
Grok 4.20xAI
24
45
o1OpenAI
22
46
Sonar Deep ResearchPerplexity AI
18

Quality scores from official model cards and published technical reports. Speed benchmarks from artificialanalysis.ai (approximate medians). Higher is better for all metrics.