Ola CEO Bhavish Aggarwal’s announcement that Krutrim would be hosting Meta’s Llama 4 models on Indian servers drew harsh criticism from multibagg.ai founder Aaditya Aanand.
Anyone with a good GPU at home could host open-source models like Llama 4, according to Aanand, who made fun of the hype.
multibagg.ai Founder Aaditya Aanand Criticizes Ola CEO Bhavish Aggarwal
Citing Krutrim’s hosting of DeepSeek models (8B to 700B parameters) and Meta’s Llama 4 Scout and Maverick, Aggarwal portrayed the launch as a significant turning point in Indian AI.
As per Aanand, “After raising 280 million dollars, he is excited for hosting an open source model? An intern having a good PC with GPU can do it in their bedroom.”
He presented it as a move toward accessible AI for Indian developers and startups, emphasizing reasonable pricing of ₹7 to ₹17 per million tokens.
Additionally, Aggarwal linked the endeavor to “technological independence” and “digital sovereignty,” stating that Indian talent can only be empowered by domestic infrastructure.
Aanand denied the pricing claims, pointing out that models such as the GPT-4o-mini and Gemini Flash are less expensive, costing less than ₹10 per million tokens.
Indian Developers Had Access to AI Models Before Krutrim Did
He mocked the idea of “democratizing AI,” asserting that Indian developers had access to these models before Krutrim did.
He attacked Aggarwal’s patriotic rhetoric, comparing it to “pretending to put the Indian flag on the moon” and claiming that hosting open-source models isn’t revolutionary.
Krutrim’s offering, according to Aanand, is merely a repackaging of pre-existing tools and lacks originality and innovation.
He claimed that Krutrim is only “renting it locally” under a patriotic marketing guise rather than creating India’s AI future.
By giving users access to robust large language models housed on Indian servers, Krutrim hopes to compete with international AI cloud platforms.
