Skip to content
InferFlux

Rankings

Models by the tokens on their cleared receipts, which anyone can count again from the chain. As of Nov 11, 20:33:34 UTC.

Leaderboard

Models by tokens, with their change against the week before
RankModelAuthorTokensAgainst the week before
1DeepSeek: DeepSeek V4.1 Flash 1017
deepseek/deepseek-v4.1-flash-1017
deepseek124KNew

Share by author

Each author's share of all tokens on cleared receipts in this window. Across every modality.

  1. deepseek100.0% 124K tokens, New

By category

From the categories an app declares for itself, in OpenRouter's four groups. Across every modality.

No app has declared a category yet, so every token is uncategorized.

Fastest models

The median output speed of each model's fastest endpoint, and its time to the first token, as the floor measured them. The floor measures on its own schedule, whatever the window above.

The floor has measured no endpoint yet.

Top apps

From this gateway's own records: an app's referer and title never reach the chain. Across every modality.

No app has named itself in this window.

How the rankings are measured

Each model figure is the tokens counted on its receipts that cleared on the chain in the window, against the window before; a receipt carries hashes and counts, never a prompt, so nothing is ranked by language or by what a request said. Apps are ranked from this gateway's records, because an app's referer and title travel with a request and not with its receipt.