Skip to main content
Meo

Median output throughput (tokens/s) — AI model leaderboard

AI models ranked by Median output throughput (tokens/s), an aggregated third-party benchmark from artificial_analysis. Higher is better. Cross-referenced against our first-party meo scores and Effective Value (𝕍).

Ranking 87 models across the full field · as of 2026-06-07.

#ModelLabMedian output throughput (tokens/s)
1Inception: Mercury 2inception1,075
2OpenAI: gpt-oss-120bopenai358
3Google: Gemini 3.1 Flash Litegoogle317
4OpenAI: gpt-oss-20bopenai268
5Arcee AI: Trinity Large Thinkingarcee-ai226
6Google: Gemini 3.5 Flashgoogle222
7OpenAI: o3 Mini Highopenai219
8OpenAI: GPT-5.1-Codex-Miniopenai219
9xAI: Grok 4.3x-ai212
10Google: Gemini 2.5 Flashgoogle197
11OpenAI: GPT-5 Codexopenai195
12StepFun: Step 3.5 Flashstepfun193
13Google: Gemini 3 Flash Previewgoogle188
14OpenAI: o3 Miniopenai187
15OpenAI: GPT-5 Nanoopenai179
16OpenAI: GPT-4.1 Nanoopenai176
17OpenAI: o4 Miniopenai173
18OpenAI: GPT-5.4 Miniopenai173
19OpenAI: GPT-5.1-Codexopenai172
20Qwen: Qwen3.6 35B A3Bqwen169
21OpenAI: GPT-5.4 Nanoopenai159
22OpenAI: o3openai157
23OpenAI: GPT-4oopenai156
24StepFun: Step 3.7 Flashstepfun148
25OpenAI: GPT-5.1openai142
26OpenAI: GPT-4o (2024-08-06)openai140
27Google: Gemini 2.5 Progoogle139
28Qwen: Qwen3.5-122B-A10Bqwen138
29OpenAI: GPT-5.2-Codexopenai136
30OpenAI: GPT-4o (2024-05-13)openai136
31Google: Gemini 3.1 Pro Previewgoogle133
32IBM: Granite 4.1 8Bibm-granite132
33Xiaomi: MiMo-V2-Flashxiaomi129
34OpenAI: GPT-4.1openai127
35inclusionAI: Ring-2.6-1Tinclusionai124
36Kwaipilot: KAT-Coder-Pro V2kwaipilot115
37Meta: Llama 4 Scoutmeta-llama113
38Meta: Llama 4 Maverickmeta-llama109
39DeepSeek: DeepSeek V4 Flashdeepseek108
40Qwen: Qwen3 Coder Nextqwen106
41Qwen: Qwen3.7 Maxqwen105
42OpenAI: GPT-4.1 Miniopenai105
43OpenAI: GPT-5 Miniopenai101
44OpenAI: GPT-5openai97
45Tencent: Hy3 previewtencent96
46OpenAI: GPT-5.3-Codexopenai95
47Reka Flash 3rekaai93
48OpenAI: GPT-5.4openai92
49Qwen: Qwen3.5-9Bqwen92
50OpenAI: GPT-5.2 Chatopenai75
51Z.ai: GLM 5.1z-ai75
52OpenAI: GPT-4o-miniopenai73
53Cohere: Command Acohere71
54Anthropic: Claude Opus 4.8anthropic71
55MiniMax: MiniMax M2.7minimax68
56Mistral: Mistral Medium 3.5mistralai66
57Qwen: Qwen3.6 27Bqwen64
58Anthropic: Claude Opus 4.7anthropic62
59OpenAI: GPT-5.5openai61
60DeepSeek: DeepSeek V4 Prodeepseek61
61Anthropic: Claude Sonnet 4.6anthropic60
62Qwen: Qwen3.6 Plusqwen53
63Qwen: Qwen3.7 Plusqwen52
64Qwen: Qwen3.5 397B A17Bqwen52
65Xiaomi: MiMo-V2.5-Proxiaomi46
66MiniMax: MiniMax M3minimax45
67OpenAI: GPT-4openai39
68Microsoft: Phi 4microsoft39
69Google: Gemma 4 31Bgoogle35
70OpenAI: GPT-4 Turboopenai33
71OpenAI: o3 Proopenai33
72Microsoft: Phi 4 Mini Instructmicrosoft24
73Z.ai: GLM 5 Turboz-ai0
74Z.ai: GLM 5V Turboz-ai0
75inclusionAI: Ling-2.6-1Tinclusionai0
76Google: Gemma 4 26B A4B (free)google0
77inclusionAI: Ling-2.6-flashinclusionai0
78Upstage: Solar Pro 3upstage0
79OpenAI: o1-proopenai0
80Prime Intellect: INTELLECT-3prime-intellect0
81Google: Gemini 2.5 Flash Lite Preview 09-2025google0
82Google: Gemma 3 27Bgoogle0
83Google: Gemma 3 12Bgoogle0
84OpenAI: GPT-5.4 Proopenai0
85OpenAI: GPT-5.5 Proopenai0
86OpenAI: o1openai0
87OpenAI: GPT-3.5 Turbo (older v0613)openai0

Artificial Analysis (artificialanalysis.ai). Redistribution requires an AA commercial license.

← All rankingsMethodology & 𝕍 →