The tension between Sarvam AI and Token Factory isn’t just a boardroom drama — it’s a fight over the foundations of how AI platforms are built, controlled, and monetised. Both companies are invested in the same underlying technology, but their interests have started to diverge as the platform matures and becomes more valuable.
At the centre of the disagreement is the question of ownership: who actually holds the rights to the core model, the training data, and the infrastructure that makes the platform work? For Sarvam AI, the value lies in the ability to steer the technical direction, optimise performance, and set the commercial terms for anyone building on top. Token Factory, on the other hand, is focused on ensuring that the tokens, APIs, and downstream integrations remain open and broadly compatible — not just controlled by a single upstream entity.
What the dispute is about
The main technical issue is control of the model weights and the data pipeline. Sarvam AI claims that its contributions to the core model entitle it to set the pace of development and restrict certain uses. Token Factory argues that the ecosystem depends on open access and that unilateral changes could break compatibility for downstream users. There’s also a dispute about who owns the data used for training: if Sarvam AI provided proprietary datasets, it may want to limit how Token Factory can use or redistribute the resulting models.
API compatibility is another sticking point. Sarvam AI may want to introduce breaking changes or proprietary extensions to its API, while Token Factory is pushing for a stable, open standard. This tension is typical when a platform shifts from early-stage collaboration to commercial competition.
What’s at stake for Sarvam AI
For Sarvam AI, the outcome determines whether it can capture more of the value created by the platform. If it retains tight control, it can charge for access, prioritise its own features, and set the terms for third-party plugins or tools. This also lets it shape the model’s roadmap, potentially locking in customers who depend on its specific implementation.
The risk is that a heavy-handed approach could fragment the ecosystem or drive developers to alternatives. If Token Factory and its allies fork the platform or create their own standard, Sarvam AI could lose network effects and face a shrinking user base.




