Sarvam AI has launched “Token Factory,” making its India-hosted inference service available to developers and enterprises. The platform lets users access several open models through a single OpenAI-compatible API, with compute supplied by GPU clusters located in India.

Both outlets describe the same core offering, but they frame it slightly differently. MarkTechPost emphasizes the service opening “to everyone” and lists the models included, while Inc42 foregrounds the idea of running global AI models on Indian servers for developers and enterprise users. Neither outlet, in the provided excerpts, describes pricing, service-level commitments, or any policy or regulatory implications.

Overall, the reporting centers on product availability and interoperability (OpenAI-compatible API) rather than changes to model capabilities themselves.