When Sarvam launched its multilingual model, the community mocked its low downloads. Now, the model has close to 900k downloads on Hugging Face and still there is nothing significant built by developers on top of it apart from some translation tools posted on X.
Same is the case for BharatGen, CoRover, or Soket AI Labs's models.
The harder question is---Why aren’t developers building for Bharat at scale on top of the models that are getting created?
The answer is---There is no incentive, no market, and no path to revenue for them. Basically, who is even buying? It is more of a philanthropic journey that developers need to take on by themselves.
The truth is, the Indian AI ecosystem rewards English benchmarks, not local language innovation. Developers focus on GPT, Mistral, and Llama—not because they don’t care about Indic languages, but because that’s what gets them jobs, funding, and visibility.
Indic AI is treated like a side project. Even national infrastructure like BHASHINI - (Digital India BHASHINI Division), though has laid the foundation well enough, hasn’t sparked a developer movement because there’s no road to monetisation for developer to start building Indic use cases.
Startups that try are met with free PoC demands or investor apathy. Until we fix the incentives, we’ll keep building wrappers, not revolutions.
The next time you see an Indian LLM drop, don’t ask how many downloads it got. Ask why the system makes building for Bharat such a thankless task.
Read full story here: https://lnkd.in/gJjgESBs
This post was originally shared by on Linkedin.