TPSTokens Per Second

AI Model Benchmarks

778 scores across 9 benchmarks — quality, coding, reasoning, and speed.

Every score we hold, including models older than 12 months. For current picks see the rankings, which use the last 12 months only.

Output throughput in tokens per second measured on standard prompts. Higher = faster responses. Source: artificialanalysis.ai and vendor-published measurements.

Model
tok/s
1
Mercury 2Inception
1007
2
Nemotron-Labs Diffusion 8BNVIDIA
865
3
DiffusionGemma 26B A4B ITGoogle DeepMind
500
4
Ling-3.0-flashInclusionai
429
5
Gemini 3.1 Flash LiteGoogle DeepMind
360
6
Gemini 3.7 FlashGoogle DeepMind
358
7
NVIDIA Nemotron 3.5 Lightning 30B-A3B (BF16)NVIDIA
321.82
8
Nemotron-3-Nano-Omni-30B-A3BNVIDIA
321
9
Gemini 3.1 Flash Lite PreviewGoogle DeepMind
319
10
Nemotron 3.5 LightningNVIDIA
318
11
Step-3.5-Flash-BaseStepFun
300
12
Step-3.5-FlashStepFun
279
13
Gemini 2.0 FlashGoogle DeepMind
228
14
Gemini 3.5 FlashGoogle DeepMind
223
15
Gemini 3 FlashGoogle DeepMind
222
16
LFM2.5-2.6BLiquid Ai
220
17
Grok 4.20xAI
215
18
LFM2.5-230MLiquid Ai
213
19
Trinity Large ThinkingArcee Ai
210
20
gpt-oss-120bOpenAI
201
21
Qwen3.7 MaxAlibaba / Qwen
200
22
Gemini 2.5 FlashGoogle DeepMind
198
23
gpt-oss-20bOpenAI
192
24
Muse Spark 1.1Meta AI
189
25
Claude Haiku 4.5Anthropic
187
26
GPT-5.4 miniOpenAI
184
27
SonarPerplexity AI
175
28
Gemma 4 26B A4B ITGoogle DeepMind
163.09
29
Bonsai 27BPrismml
163
30
Qwen3.5-35B-A3BAlibaba / Qwen
162
31
Mistral Small 4Mistral AI
161
32
Llama 4 ScoutMeta AI
160
33
Hy3 PreviewTencent
158
34
GPT-5.4 NanoOpenAI
155
35
Mistral Medium 3.5Mistral AI
152
36
Mistral SabaMistral AI
150
37
Gemma 4 12BGoogle DeepMind
147
38
Grok 4 minixAI
145
39
NVIDIA Nemotron-3-Super-120B-A12BNVIDIA
142
40
Qwen3.6 35B A3BAlibaba / Qwen
140
41
GPT-5.6 LunaOpenAI
138
42
Grok 3 minixAI
138
43
GPT-5.4OpenAI
136
44
Nex-N2-ProNex Agi
135
45
Granite 4.1 8BIbm
125
46
GPT-5.3-CodexOpenAI
125
47
DeepSeek V4 FlashDeepSeek
122
48
Ring-2.6-1TInclusionai
122
49
Sonar ReasoningPerplexity AI
120
50
Step-3.7-FlashStepFun
115
51
o4-miniOpenAI
115
52
MiniMax-M2MiniMax
114
53
GLM-4.7Zhipu AI
114
54
GPT-5.6 TerraOpenAI
114
55
Grok 4.3xAI
113
56
GPT-4.1OpenAI
112
57
Kimi K2 ThinkingMoonshot AI
108
58
o3-miniOpenAI
108
59
MiniMax-M3MiniMax
108
60
Muse GlimmerMeta AI
106
61
Ornith-1.0Deepreinforce
103
62
KAT-Coder-Pro V2Kwaipilot
102
63
GPT-4oOpenAI
98
64
Sonar ProPerplexity AI
95
65
GPT-5.5OpenAI
92
66
MiniMax-M2.5MiniMax
91
67
Llama 4 MaverickMeta AI
90
68
GLM-5.2Zhipu AI
87
69
Ling-2.6-flashInclusionai
87
70
Claude Sonnet 5Anthropic
87
71
Claude Opus 4.5Anthropic
83
72
Qwen3.5-27BAlibaba / Qwen
82
73
Claude Sonnet 4.5Anthropic
82
74
GLM-5.3Zhipu AI
79
75
Claude 3.5 SonnetAnthropic
78
76
GPT-5.2OpenAI
78
77
Qwen3.5 397B A17BAlibaba / Qwen
78
78
Hy3Tencent
76
79
GPT-5.6 SolOpenAI
75
80
Mistral Medium 3Mistral AI
75
81
Mistral Large 2Mistral AI
72
82
InklingThinking Machines
71
83
Qwen3-MaxAlibaba / Qwen
70
84
GLM-4.6Zhipu AI
70
85
DeepSeek V3DeepSeek
70
86
Claude Opus 4.6Anthropic
70
87
Inkling SmallThinking Machines
69
88
Claude Fable 5Anthropic
68
89
Claude 3.7 SonnetAnthropic
68
90
DeepSeek V4 ProDeepSeek
68
91
Claude Sonnet 4.6Anthropic
66
92
Sonar Reasoning ProPerplexity AI
65
93
Qwen3 72BAlibaba / Qwen
65
94
Gemini 2.5 ProGoogle DeepMind
62
95
GLM 5Zhipu AI
62
96
Grok 4.6xAI
62
97
Claude Opus 4.7Anthropic
61
98
Claude Opus 4.8Anthropic
60
99
Phi-4Microsoft
60
100
MiniMax M2.7MiniMax
60
101
Qwen3.6 27BAlibaba / Qwen
59
102
Mistral Large 3Mistral AI
59
103
MiMo-V2.5-ProXiaomi
58
104
DeepSeek R1DeepSeek
58
105
Grok 4.5xAI
57
106
Kimi K2.6Moonshot AI
57
107
MiniMax M2.7MiniMax
57
108
Claude Opus 5Anthropic
56
109
Qwen3.7 PlusAlibaba / Qwen
56
110
Grok 3xAI
55
111
Qwen3.5-OmniAlibaba / Qwen
51
112
QwQ-32BAlibaba / Qwen
50
113
Mistral Medium 3Mistral AI
49
114
Devstral 2Mistral AI
45
115
Qwen3.8-2.4T-A95BAlibaba / Qwen
45
116
Gemini 1.5 ProGoogle DeepMind
45
117
Qwen3.8-MaxAlibaba / Qwen
45
118
Muse Glimmer 30BMeta AI
43.167
119
GPT-4.5OpenAI
42
120
Kimi K2.5Moonshot AI
41
121
Claude Opus 4Anthropic
40
122
Kimi K2 0905Moonshot AI
39
123
LongCat 2.0Meituan
39
124
Kimi K3Moonshot AI
39
125
Kimi K2.7 CodeMoonshot AI
38
126
Gemma 4 31BGoogle DeepMind
35
127
o3OpenAI
32
128
Qwen3.5-4BAlibaba / Qwen
29
129
GLM-5.1Zhipu AI
28
130
Grok 4xAI
28
131
North Mini CodeCohere
27
132
Llama 3.1 405BMeta AI
26
133
o1OpenAI
22
134
Sonar Deep ResearchPerplexity AI
18

Quality scores from official model cards, published technical reports, and independent leaderboards (vals.ai, Vectara). Speed benchmarks from artificialanalysis.ai (approximate medians). Higher is better for all metrics except Hallucination Rate, where lower is better.