For the first time since 2024, developers on the OpenRouter platform spent more on OpenAI models than on models from Anthropic. OpenRouter is a platform that allows developers to access AI models from different companies through a single API. Spending on the platform provides an indication of how widely different AI models are being used.
According to OpenRouter’s head of analytics, OpenAI’s new GPT-6 Astra accounted for 19% of total platform spending last week. Anthropic’s Claude Fable 5.1 accounted for 6%. The two models have the same pricing on OpenRouter: $10 per 1 million input tokens and $50 per 1 million output tokens.
OpenAI introduced GPT-6 Astra on September 3. The model is designed for tasks including long-context processing, coding and computer use. In benchmarks published by OpenAI, Astra achieved higher results than the previous GPT-5.6 Sol on some tasks. For example, it scored 96.3% on the MRCR long-context test in the 512,000-to-1-million-token range.
However, Astra did not lead in every benchmark. Claude Fable 5.1 recorded a higher result on the Artificial Analysis Intelligence Index. Performance therefore varies depending on the task.
According to OpenRouter data, since 2024, developers on the platform had spent more on Anthropic models than on OpenAI models. That changed following the release of GPT-6 Astra. In one week of spending, Astra alone accounted for 19%, while Fable 5.1 stood at 6%.
GPT-6 Astra supports a context window of 1.05 million tokens, while Fable 5.1 supports up to 1 million tokens. For now, the figures cover only a one-week period. Spending over the coming weeks will show whether the increase in demand for OpenAI models remains sustained on OpenRouter.
