VerdictPal
Join

Benchmarks / Performance

AA output speed

Output tokens per second

Median output tokens per second after the first chunk arrives, measured on each model's default API provider with a ~1k-token input prompt (AA medium workload).

Artificial AnalysisPerformancePrimary · PerformanceFlagshipActiveUnknown contaminationSince 2026
What this does not measure
  • IDE or chat-app latency — only first-party API routes AA tracks as the default provider.
  • Quality or reasoning depth — speed-only slice. A fast model can still fail hard tasks.
Analysis

Why this benchmark is useful

When you care how fast tokens stream after the first chunk — IDE feel, chat responsiveness, batch throughput — not whether the answer is right.

Scope

Coverage map

Task family
Performance
Format
Streaming generation after first token; median over the past 72 hours on AA's harness.
Scoring
Higher tokens/s is faster. AA uses OpenAI tiktoken o200k_base for token counts.
Maintainer
Artificial Analysis
Reading guide

How to read the scores

Higher median tokens/s is faster. AA pins each model's default API provider and a ~1k-token medium workload; IDE or alternate providers can diverge.

Blind spots

What it does not cover

  • IDE or chat-app latency — only first-party API routes AA tracks as the default provider.
  • Quality or reasoning depth — speed-only slice. A fast model can still fail hard tasks.
Scores

Evidence ledger

195 rows
195195 rows
414 t/sbest score
195source-checked
1sources
2026-10-12to 2026-09-05
195

api · api

Distribution

Where the rows land

0255075100

Normalized to this benchmark's axis (0–100). Open the table for raw units.

Timeline

Newest receipts

  1. Gemini 3.5 Flash-Lite414 t/s ·
  2. Gemini 3.1 Flash Lite361 t/s ·
  3. Ling 3.0 Flash347 t/s ·
  4. Gemini 3.7 Flash314 t/s ·
  5. Nemotron 3.5 Lightning301 t/s ·

Scores use this benchmark's own unit and axis, not a universal quality score.

#ModelRelease dateScoreProvenanceTrust
1Gemini 3.5 Flash-LiteGoogle2026-07-21414 t/s
2026-10-12
Source-checked
2Gemini 3.1 Flash LiteGoogle2026-05-07361 t/s
2026-10-12
Source-checked
3Ling 3.0 FlashOther2026-08-04347 t/s
2026-10-12
Source-checked
4Gemini 3.7 FlashGoogle2026-08-13314 t/s
2026-10-12
Source-checked
5Nemotron 3.5 LightningNVIDIA2026-08-11301 t/s
2026-10-12
Source-checked
6Muse Spark 1.2Meta2026-08-05272 t/s
2026-10-12
Source-checked
7Nemotron 3 Nano Omni 30B A3B ReasoningNVIDIA2026-04-29243 t/s
2026-10-12
Source-checked
8GPT-5.4 MiniOpenAI2026-03-17236 t/s
2026-10-12
Source-checked
9Claude Haiku 5.5Anthropic2026-10-07235 t/s
2026-10-12
Source-checked
10Granite 4.2 3BOther2026-08-25211 t/s
2026-10-12
Source-checked
11DeepSeek V4.1 FlashDeepSeek2026-09-10205 t/s
2026-10-12
Source-checked
12DeepSeek V4 Flash 0731 (Reasoning, Max Effort)DeepSeek2026-07-31202 t/s
2026-10-12
Source-checked
13Gemini 3 Flash PreviewGoogle2025-12-17201 t/s
2026-10-12
Source-checked
14DeepSeek V4 Flash Vision (Reasoning, Max Effort)DeepSeek2026-08-21199 t/s
2026-10-12
Source-checked
15Gemini 3.5 Flash (medium)Google2026-05-19198 t/s
2026-10-12
Source-checked
16Qwen3.7 MaxAlibaba2026-05-19197 t/s
2026-10-12
Source-checked
17Inkling SmallOther2026-07-30193 t/s
2026-10-12
Source-checked
18Muse Spark 1.3Meta2026-09-02190 t/s
2026-09-05
Source-checked
19Gemini 3.5 FlashGoogle2026-05-19189 t/s
2026-10-12
Source-checked
20Gemini 3 Flash Preview (Reasoning)Google2025-12-17188 t/s
2026-10-12
Source-checked
21Inkling (xhigh)Other2026-07-15187 t/s
2026-10-12
Source-checked
22Gemini 3.6 Flash (high)Google2026-07-21186 t/s
2026-10-12
Source-checked
23Gemini 3.5 Flash (minimal)Google2026-05-19181 t/s
2026-10-12
Source-checked
24GPT-5.4 NanoOpenAI2026-03-17172 t/s
2026-10-12
Source-checked
25GPT-5 Codex (high)OpenAI2025-09-23169 t/s
2026-10-12
Source-checked
26Mistral Medium 3.5Mistral2026-04-29163 t/s
2026-10-12
Source-checked
27Kimi K2 ThinkingMoonshot2025-11-06161 t/s
2026-10-12
Source-checked
28Step 3.7 FlashStepFun2026-06-01160 t/s
2026-10-12
Source-checked
29GLM-5.2 (Non-reasoning)Zhipu2026-06-16160 t/s
2026-10-12
Source-checked
30Nemotron 3 Ultra 550B A55B (Reasoning)NVIDIA2026-06-04155 t/s
2026-10-12
Source-checked
31GPT-5.5 Instant (June 2026)OpenAI2026-06-25153 t/s
2026-10-12
Source-checked
32Qwen3.6 35B A3B (Non-reasoning)Alibaba2026-04-16137 t/s
2026-10-12
Source-checked
33GPT-6 LunaOpenAI2026-09-22136 t/s
2026-10-12
Source-checked
34Claude Sonnet 5.5Anthropic2026-09-28135 t/s
2026-10-12
Source-checked
35Gemini 3.8 FlashGoogle2026-09-02132 t/s
2026-10-12
Source-checked
36Gemini 3.1 Pro PreviewGoogle2026-02-19131 t/s
2026-10-12
Source-checked
37Muse GlimmerMeta2026-08-10131 t/s
2026-10-12
Source-checked
38GPT-5.2 Codex (xhigh)OpenAI2025-12-11129 t/s
2026-10-12
Source-checked
39Qwen3.6 35B A3B (Reasoning)Alibaba2026-04-16129 t/s
2026-10-12
Source-checked
40OpenAI o3OpenAI2025-04-16129 t/s
2026-10-12
Source-checked
41Qwen3.5 122B A10B (Reasoning)Alibaba2026-02-24129 t/s
2026-10-12
Source-checked
42GPT-5.1 Codex (high)OpenAI2025-11-13124 t/s
2026-10-12
Source-checked
43GPT-6 SolOpenAI2026-09-22122 t/s
2026-10-12
Source-checked
44GPT-5.6 TerraOpenAI2026-07-09122 t/s
2026-10-12
Source-checked
45GPT-5.6 LunaOpenAI2026-07-09120 t/s
2026-10-12
Source-checked
46K2 Horizon 375B A23BOther2026-09-03119 t/s
2026-10-12
Source-checked
47North Mini CodeCohere2026-06-09119 t/s
2026-10-12
Source-checked
48GPT-5.1OpenAI2025-11-13119 t/s
2026-10-12
Source-checked
49Ring-2.6-1TOther2026-05-08118 t/s
2026-10-12
Source-checked
50GPT-5.3 Codex (xhigh)OpenAI2026-02-05115 t/s
2026-10-12
Source-checked
51Gemma 4 12B (Non-reasoning)Google2026-06-10114 t/s
2026-10-12
Source-checked
52Grok 4.3 (Non-reasoning)xAI2026-04-30113 t/s
2026-10-12
Source-checked
53Gemma 4 12B (Reasoning)Google2026-06-07111 t/s
2026-10-12
Source-checked
54MiniMax M3MiniMax2026-05-31107 t/s
2026-10-12
Source-checked
55GPT-5 (high)OpenAI2025-08-07104 t/s
2026-10-12
Source-checked
56GPT-5 (low)OpenAI2025-08-07104 t/s
2026-10-12
Source-checked
57Grok 4.3xAI2026-04-30103 t/s
2026-10-12
Source-checked
58GPT-5 (medium)OpenAI2025-08-07103 t/s
2026-10-12
Source-checked
59Muse Spark 1.1Meta2026-07-09102 t/s
2026-10-12
Source-checked
60Claude Opus 5.5Anthropic2026-09-2296 t/s
2026-10-12
Source-checked
61Solar Pro 4Other2026-08-0695 t/s
2026-10-12
Source-checked
62Grok 4.3 (low)xAI2026-04-3095 t/s
2026-10-12
Source-checked
63GPT 5.5OpenAI2026-04-2394 t/s
2026-10-12
Source-checked
64GPT 5.4OpenAI2026-03-0594 t/s
2026-10-12
Source-checked
65Grok 4.3 (medium)xAI2026-04-3094 t/s
2026-10-12
Source-checked
66DeepSeek V4 Pro (Non-reasoning)DeepSeek2026-04-2492 t/s
2026-10-12
Source-checked
67DeepSeek V4 Pro 0813 (Reasoning, Max Effort)DeepSeek2026-08-1392 t/s
2026-10-12
Source-checked
68Grok 4.20 ReasoningxAI2026-03-0590 t/s
2026-10-12
Source-checked
69GPT-5.5 (high)OpenAI2026-04-2389 t/s
2026-10-12
Source-checked
70GLM-5.2Zhipu2026-06-1388 t/s
2026-10-12
Source-checked
71Qwen3.5 397B A17B (Reasoning)Alibaba2026-02-1688 t/s
2026-10-12
Source-checked
72GPT-5.5 (medium)OpenAI2026-04-2387 t/s
2026-10-12
Source-checked
73GPT-5.6 SolOpenAI2026-07-0985 t/s
2026-10-12
Source-checked
74Grok 4.20 0309 v2 (Non-reasoning)xAI2026-04-0785 t/s
2026-10-12
Source-checked
75Hy3Other2026-07-0685 t/s
2026-10-12
Source-checked
76Qwen3.5 397B A17B (Non-reasoning)Alibaba2026-02-1684 t/s
2026-10-12
Source-checked
77GPT-5.5 (Non-reasoning)OpenAI2026-04-2384 t/s
2026-10-12
Source-checked
78Grok 4.7xAI2026-09-2184 t/s
2026-10-12
Source-checked
79GPT-5.5 (low)OpenAI2026-04-2384 t/s
2026-10-12
Source-checked
80GPT-5.2OpenAI2025-12-1181 t/s
2026-10-12
Source-checked
81Claude Sonnet 5Anthropic2026-06-3079 t/s
2026-10-12
Source-checked
82Granite 4.2 30BOther2026-08-2577 t/s
2026-10-12
Source-checked
83Qwen3.5 27B (Reasoning)Alibaba2026-02-2476 t/s
2026-10-12
Source-checked
84GLM-5 (Reasoning)Zhipu2026-02-1175 t/s
2026-10-12
Source-checked
85Mistral Large 3Mistral2025-12-0274 t/s
2026-10-12
Source-checked
86Grok 4xAI2025-07-1073 t/s
2026-10-12
Source-checked
87Kimi K2.7 CodeMoonshot2026-06-1270 t/s
2026-10-12
Source-checked
88MiniMax M2.5MiniMax2026-04-0168 t/s
2026-10-12
Source-checked
89Llama 4 ScoutMeta2026-04-0567 t/s
2026-10-12
Source-checked
90MiniMax-M2.1MiniMax2025-12-2367 t/s
2026-10-12
Source-checked
91Claude Fable 5.1Anthropic2026-09-0167 t/s
2026-10-12
Source-checked
92GLM-5.3Zhipu2026-08-1466 t/s
2026-10-12
Source-checked
93GPT-6.1 SolOpenAI2026-09-2966 t/s
2026-10-12
Source-checked
94Grok Build 0.1 0616xAI2026-06-1665 t/s
2026-10-12
Source-checked
95LongCat-2.0Meituan2026-06-2965 t/s
2026-10-12
Source-checked
96Claude Fable 5Anthropic2026-06-0965 t/s
2026-10-12
Source-checked
97GPT-6 AstraOpenAI2026-09-0364 t/s
2026-09-05
Source-checked
98MiniMax M2.7MiniMax2026-04-1564 t/s
2026-10-12
Source-checked
99GLM-5 (Non-reasoning)Zhipu2026-02-1163 t/s
2026-10-12
Source-checked
100GLM-5.3-FlashZhipu2026-08-2661 t/s
2026-10-12
Source-checked
101GLM-5.1Zhipu2026-04-0760 t/s
2026-10-12
Source-checked
102Claude Sonnet 4.6 (Adaptive Reasoning, Max Effort)Anthropic2026-02-1760 t/s
2026-10-12
Source-checked
103GPT-5.4 (low)OpenAI2026-03-0560 t/s
2026-10-12
Source-checked
104Command ACohere2026-03-1560 t/s
2026-10-12
Source-checked
105Granite 4.2 8BOther2026-08-2559 t/s
2026-10-12
Source-checked
106Claude Opus 4.8Anthropic2026-05-2858 t/s
2026-10-12
Source-checked
107Kimi K2.5Moonshot2026-01-2757 t/s
2026-10-12
Source-checked
108Grok 4.5xAI2026-07-0857 t/s
2026-10-12
Source-checked
109Grok 4.6xAI2026-08-1257 t/s
2026-10-12
Source-checked
110Qwen3.6 27B (Non-reasoning)Alibaba2026-04-2256 t/s
2026-10-12
Source-checked
111Qwen3.6 PlusAlibaba2026-04-0256 t/s
2026-10-12
Source-checked
112Qwen 3.7 PlusAlibaba2026-06-0255 t/s
2026-10-12
Source-checked
113Qwen3.6 27B (Reasoning)Alibaba2026-04-2255 t/s
2026-10-12
Source-checked
114Claude Opus 5 (Adaptive Reasoning, Max Effort)Anthropic2026-07-2454 t/s
2026-10-12
Source-checked
115Claude Opus 5 (Adaptive Reasoning, Low Effort)Anthropic2026-07-2454 t/s
2026-10-12
Source-checked
116Claude Opus 5 (Adaptive Reasoning, High Effort)Anthropic2026-07-2453 t/s
2026-10-12
Source-checked
117Claude Opus 5 (Adaptive Reasoning, Xhigh Effort)Anthropic2026-07-2453 t/s
2026-10-12
Source-checked
118Claude Opus 4.5 (Reasoning)Anthropic2025-11-2453 t/s
2026-10-12
Source-checked
119Claude Opus 5 (Adaptive Reasoning, Medium Effort)Anthropic2026-07-2452 t/s
2026-10-12
Source-checked
120Qwen3.8-Flash-NextAlibaba2026-08-2651 t/s
2026-10-12
Source-checked
121Llama 4 MaverickMeta2026-04-0550 t/s
2026-10-12
Source-checked
122Claude Opus 4.7Anthropic2026-04-1650 t/s
2026-10-12
Source-checked
123MiMo-V2.6-FlashXiaomi2026-09-2249 t/s
2026-10-12
Source-checked
124Qwen3.8 27BAlibaba2026-08-1446 t/s
2026-10-12
Source-checked
125Claude Opus 4.5Anthropic2025-11-2446 t/s
2026-10-12
Source-checked
126Claude 4.5 Sonnet (Reasoning)Anthropic2025-09-2945 t/s
2026-10-12
Source-checked
127Claude Sonnet 4.6Anthropic2026-02-1744 t/s
2026-10-12
Source-checked
128Claude Opus 4.7 (Non-reasoning, High Effort)Anthropic2026-04-1644 t/s
2026-10-12
Source-checked
129GLM-4.7 (Reasoning)Zhipu2025-12-2244 t/s
2026-10-12
Source-checked
130Kimi K2.6Moonshot2026-04-2043 t/s
2026-10-12
Source-checked
131Kimi K3 (low)Moonshot2026-07-1643 t/s
2026-10-12
Source-checked
132Claude Sonnet 4.6 (Non-reasoning, Low Effort)Anthropic2026-02-1742 t/s
2026-10-12
Source-checked
133Kimi K2.6 (Non-reasoning)Other2026-04-2042 t/s
2026-10-12
Source-checked
134Kimi K3Moonshot2026-07-1642 t/s
2026-10-12
Source-checked
135MiMo-V2.6-ProXiaomi2026-09-2240 t/s
2026-10-12
Source-checked
136Claude Opus 4.6 (Adaptive Reasoning, Max Effort)Anthropic2026-02-0540 t/s
2026-10-12
Source-checked
137Claude Opus 4.6Anthropic2026-02-0538 t/s
2026-10-12
Source-checked
138Qwen3.8 2.4T A95BAlibaba2026-08-1237 t/s
2026-10-12
Source-checked
139MiMo-V2.5Xiaomi2026-04-2236 t/s
2026-10-12
Source-checked
140Qwen3.8 MaxAlibaba2026-08-0336 t/s
2026-10-12
Source-checked
141Gemma 4 31B ITGoogle2026-03-0135 t/s
2026-10-12
Source-checked
142MiMo-V2.5-Pro (Non-reasoning)Xiaomi2026-04-2234 t/s
2026-10-12
Source-checked
143DeepSeek R1DeepSeek2025-01-2031 t/s
2026-10-12
Source-checked
144MiMo-V2.5-ProXiaomi2026-04-2230 t/s
2026-10-12
Source-checked
145Ling 3.0 TinyOther2026-08-0630 t/s
2026-10-12
Source-checked
146A.X-K2Other2026-08-120 t/s
2026-09-28
Source-checked
147Apodex 1.1Other2026-08-300 t/s
2026-09-28
Source-checked
148Claude 3 OpusAnthropic2024-03-040 t/s
2026-09-28
Source-checked
149Claude 4 Opus (Reasoning)Anthropic2025-05-220 t/s
2026-09-28
Source-checked
150Claude 4.1 Opus (Reasoning)Anthropic2025-08-050 t/s
2026-09-28
Source-checked
151DeepSeek Coder V2DeepSeek2024-05-150 t/s
2026-09-28
Source-checked
152DeepSeek V3.2 (Reasoning)DeepSeek2025-12-010 t/s
2026-09-28
Source-checked
153DiffusionGemma 26B A4BGoogle2026-06-100 t/s
2026-09-28
Source-checked
154EXAONE 4.5 33BOther2026-04-090 t/s
2026-09-28
Source-checked
155G9v3-39A5BOther2026-08-030 t/s
2026-09-28
Source-checked
156G9v3-3BOther2026-07-230 t/s
2026-09-28
Source-checked
157Gemini 1.0 UltraGoogle2024-02-080 t/s
2026-09-28
Source-checked
158Gemini 3 ProGoogle2025-11-180 t/s
2026-09-28
Source-checked
159Gemini 3 Pro Preview (low)Google2025-11-180 t/s
2026-09-28
Source-checked
160GLM 5V Turbo (Reasoning)Zhipu2026-04-010 t/s
2026-09-28
Source-checked
161GLM-5-TurboZhipu2026-03-150 t/s
2026-09-28
Source-checked
162GPT-5.2 (medium)OpenAI2025-12-110 t/s
2026-09-28
Source-checked
163GPT-5.4 ProOpenAI2026-03-050 t/s
2026-09-28
Source-checked
164GPT-5.5 Instant (May 2026)OpenAI2026-05-050 t/s
2026-09-28
Source-checked
165GPT-5.5 ProOpenAI2026-04-230 t/s
2026-09-28
Source-checked
166Granite 4.1 30BOther2026-04-290 t/s
2026-09-28
Source-checked
167Grok 4.20 0309 (Reasoning)xAI2026-03-100 t/s
2026-09-28
Source-checked
168Hy3-preview (Non-reasoning)Other2026-04-230 t/s
2026-09-28
Source-checked
169Hy3-preview (Reasoning)Other2026-04-230 t/s
2026-09-28
Source-checked
170JT-35B-FlashOther2026-05-140 t/s
2026-09-28
Source-checked
171JT-4.1 Flash 236B A21BOther2026-07-090 t/s
2026-09-28
Source-checked
172JT-MINIOther2026-04-150 t/s
2026-09-28
Source-checked
173K-EXAONE 2.0 0803Other2026-08-120 t/s
2026-09-28
Source-checked
174KAT Coder Pro V2Other2026-03-270 t/s
2026-09-28
Source-checked
175LFM2.5-8B-A1BLiquid2026-06-080 t/s
2026-09-28
Source-checked
176Ling 2.6 FlashOther2026-04-210 t/s
2026-09-28
Source-checked
177Ling-2.6-1TOther2026-04-230 t/s
2026-09-28
Source-checked
178MiMo-V2-Flash (Feb 2026)Xiaomi2025-12-160 t/s
2026-09-28
Source-checked
179MiMo-V2-Flash (Reasoning)Xiaomi2025-12-160 t/s
2026-09-28
Source-checked
180MiMo-V2-OmniXiaomi2026-03-190 t/s
2026-09-28
Source-checked
181MiMo-V2-Omni-0327Xiaomi2026-03-270 t/s
2026-09-28
Source-checked
182MiMo-V2-ProXiaomi2026-03-180 t/s
2026-09-28
Source-checked
183MiniCPM-V 4.6 1.3BOpenBMB2026-05-110 t/s
2026-09-28
Source-checked
184MiniCPM5-1B (Non-reasoning)OpenBMB2026-05-250 t/s
2026-09-28
Source-checked
185MiniCPM5-1B (Reasoning)OpenBMB2026-06-040 t/s
2026-09-28
Source-checked
186MiniCPM5-2BOpenBMB2026-09-070 t/s
2026-09-28
Source-checked
187Motif 3Other2026-08-120 t/s
2026-09-28
Source-checked
188Motif 3 (Beta)Other2026-07-210 t/s
2026-09-28
Source-checked
189Muse SparkOther2026-01-010 t/s
2026-09-28
Source-checked
190Nex-N2-ProOther2026-06-020 t/s
2026-09-28
Source-checked
191o3-proOpenAI2025-06-100 t/s
2026-09-28
Source-checked
192OpenAI o1OpenAI2024-09-120 t/s
2026-09-28
Source-checked
193Qwen3 Max ThinkingAlibaba2026-01-260 t/s
2026-09-28
Source-checked
194Qwen3.6 Max PreviewAlibaba2026-04-200 t/s
2026-09-28
Source-checked
195Solar Open2 250BOther2026-08-120 t/s
2026-09-28
Source-checked
Method

What it covers

This benchmark sits in the performance family. It uses Streaming generation after first token; median over the past 72 hours on AA's harness. The score should travel with its task format, scoring method, source date, and benchmark version.

Score ceiling

Where it breaks down

No ceiling note is recorded yet. Treat clustering near the top as a warning that the benchmark may no longer separate frontier models.

Tools that report it

Receipts

Sources and further reading

← All benchmarksHow we score

Questions about this benchmark

Every answer below is assembled from the dated fields on this page. Nothing is written separately for search.

What does AA output speed measure?

Median output tokens per second after the first chunk arrives, measured on each model's default API provider with a ~1k-token input prompt (AA medium workload).

What does a high AA output speed score not prove?

A strong AA output speed result says nothing about:

  • IDE or chat-app latency — only first-party API routes AA tracks as the default provider.
  • Quality or reasoning depth — speed-only slice. A fast model can still fail hard tasks.

How is AA output speed scored?

Higher tokens/s is faster. AA uses OpenAI tiktoken o200k_base for token counts. Task format: Streaming generation after first token; median over the past 72 hours on AA's harness.

Is AA output speed saturated?

AA output speed is currently marked Active in the atlas.

Can AA output speed results be contaminated by training data?

Contamination risk for AA output speed is graded Unknown contamination. Treat every row on this page as a public claim with a source and a date, not as a controlled experiment.

The Pack · Editorial newsletter

Know when the rules move.

One short email when a university changes its AI rule for theses, or a research tool changes its pricing or privacy terms. No tracking, no third-party analytics. Unsubscribe in one click.

The thesis AI checklist is free to use, no signup needed →