Fable 5.2 launch week: 11% of Anthropic spend on OpenRouter
- Source
- x.com
- Date
After 1 full week, Fable 5.2 sits at 11% of Anthropic model spend. That is a nearly identical launch week figure compared with Fable 5.1 (12%) and Fable 5 (13%). Opus 5.5 continues to lead in Anthropic spend with 46% wallet share across Anthropic's family of models

Context
By OpenRouter's own yardstick, Claude Fable 5.2 had an ordinary first week. OpenRouter, a service that lets developers call models from many providers through one API, reported that one full week after launch, Fable 5.2 accounted for 11% of the money spent on Anthropic models through the platform. It compared that with the first weeks of the two earlier Fable releases, 12% for Fable 5.1 and 13% for Fable 5, and called the three figures nearly identical. Anthropic announced Fable 5.2 on November 9, so the new model has started about where its predecessors did.
The same post says Opus 5.5 still leads Anthropic spend on OpenRouter with 46%, continuing the position OpenRouter described in October, when Opus 5.5 was the top model by spend across the whole platform. These shares measure money, not requests or , and the two models are priced differently. OpenRouter's model pages list Fable 5.2 at $8 per million input tokens and $40 per million output tokens with most providers (Fable 5.2), twice the $4 and $20 listed for Opus 5.5 (Opus 5.5). A Fable 5.2 request therefore adds more to its spend share than the same request sent to Opus 5.5, and the post does not say how much work either model actually handled.
The figures also describe one channel. They count spending routed through OpenRouter, not developers who call Anthropic, Amazon Bedrock or Google Vertex directly. The post supports a narrow reading: on OpenRouter, Fable 5.2's uptake in its first week matched the pattern of earlier Fable launches, and Opus 5.5 remained the Anthropic model customers there spent the most on. It does not show how Fable 5.2 is doing with Anthropic's customers as a whole.
- token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.
Checking sign-in…
Loading comments…





