Sources
The anonymous OpenRouter model "Ox Alpha" that served 42 trillion tokens in six days turned out to be Zhipu's GLM-5.3-Flash
Fireship's 2026-09-01 video resolves the stealth-model mystery that dominated OpenRouter for a week: Ox Alpha, which moved 42 trillion tokens in six days, was Zhipu's GLM-5.3-Flash, priced roughly 40x below Claude. The volume figure is the interesting part for builders, because it means a large share of agentic token traffic silently rerouted to a Chinese open-weights model on price alone before anyone knew what it was. Matthew Berman covered the same unmasking on 2026-08-29 under the title "Cancel your subscriptions, Ox-Alpha is here."
Source
↳ Follow the thread