Skip to content
Refreshed daily

AI Router Traffic Statistics

A live picture of what actually flows through our gateway — token volume, model mix, provider market share, latency percentiles and workload categories. Refreshed every day.

Tokens processed
105.87B
+71.5%vs previous period
API requests
20.8M
+66.9%vs previous period
Active models
15
Active providers
14

Daily token volume

Aggregate across all models and providers

Provider market share

Share of total token traffic by provider

  • anthropic31.6%
  • google26.4%
  • openai22.7%
  • deepseek7.9%
  • mistral4.4%
  • cerebras2.8%
  • xai2.1%
  • moonshot2.0%

Workload categories

How our customers actually use the gateway

  • Programming65.6% · 14.84B
  • General24.4% · 5.52B
  • Reasoning6.2% · 1.41B
  • Agents / Tool use3.8% · 865.69M

Top models

Ranked by token volume in the selected window

#ModelProviderRequestsTokensShare
1
Claude Sonnet 4.6
anthropic630K4.74B23.7%
2
Gemini 3 Pro
google374K3.08B15.4%
3
GPT-5.4
openai497K2.24B11.2%
4
Gemini 3 Flash
google319K1.99B10.0%
5
GPT-5.5
openai310K1.46B7.3%
6
Claude Opus 4.7
anthropic245K1.45B7.3%
7
DeepSeek V3.2
deepseek230K1.06B5.3%
8
DeepSeek R1.1
deepseek115K612.02M3.1%
9
Mistral Large 3
mistral102K553.37M2.8%
10
o4-mini
openai109K521.48M2.6%
11
Gemini 3 Flash-Lite
google167K512.80M2.6%
12
Claude Haiku 4.5
anthropic128K507.36M2.5%
13
Kimi K2
moonshot66K423.43M2.1%
14
Llama 4 Scout (Cerebras)
cerebras128K421.75M2.1%
15
GPT-5.4 Mini
openai152K391.27M2.0%

Provider response times

Time-to-first-token (ms) percentiles

Providerp50p90p99Tokens/sec
cerebras
1202203811730
groq
177315551562
deepinfra
4067721486157
mistral
410762148595
fireworks
4147841504181
google
4598691641139
together
4678711657164
openai
5019251755102
cohere
56010161888102
xai
5831125212790
deepseek
6561235223862
anthropic
7991455270680
moonshot
9161583254161
perplexity
9331699295865

All values are aggregated across the gateway. Individual request content is never exposed. Refreshed every 30 minutes.