Published model tokens by complete UTC settlement day
Day (UTC)
Model
Tokens
Requests
2026-08-21
No published usage
0
0
2026-08-22
No published usage
0
0
2026-08-23
No published usage
0
0
2026-08-24
No published usage
0
0
2026-08-25
No published usage
0
0
2026-08-26
No published usage
0
0
2026-08-27
No published usage
0
0
2026-08-28
No published usage
0
0
2026-08-29
No published usage
0
0
2026-08-30
No published usage
0
0
2026-08-31
No published usage
0
0
2026-09-01
No published usage
0
0
2026-09-02
No published usage
0
0
2026-09-03
No published usage
0
0
2026-09-04
No published usage
0
0
2026-09-05
No published usage
0
0
2026-09-06
No published usage
0
0
2026-09-07
No published usage
0
0
2026-09-08
Gemini 3.8 Flash
53,579
10
2026-09-08
MiMo M2.5
29,109
16
2026-09-09
No published usage
0
0
2026-09-10
GPT 5.6 Luna
1,754,130
1,188
2026-09-11
No published usage
0
0
2026-09-12
No published usage
0
0
2026-09-13
No published usage
0
0
2026-09-14
MiMo M2.5
79,800
36
2026-09-15
DeepSeek V4.1 Flash
10,057,870
285
2026-09-15
GPT 5.6 Sol
2,932,355
1,123
2026-09-15
MiMo M2.5
89,937
41
2026-09-16
DeepSeek V4.1 Flash
52,777,235
922
2026-09-17
Nova Micro
5,181,388
850
2026-09-18
No published usage
0
0
2026-09-19
No published usage
0
0
Only eligible usage is shown. Shares use published token totals.
How we count
We count successful, settled language-model requests with provider-reported usage. Tokens include cached input, recorded audio input and reasoning; image/tool counts, estimates, failures and calls without settled usage records are excluded.
A model needs 10 eligible requests from three accounts in a complete UTC day. Opted-out, unlinked and excluded internal/test accounts are omitted. Below-threshold usage contributes to no chart or total, including Others.
Days follow settlement time; late settlements can appear later. Changes compare equal previous periods and need complete history and a published baseline. Daily values are fixed when published; later arrivals or privacy choices do not rewrite them.
These are MiniRouter adoption rankings. Tokenization, verbosity and automatic routing influence the mix. Small-group thresholds do not guarantee anonymity. Privacy policy →