Ola Krutrim Makes DeepSeek Available At Introductory Pricing Of Rs. 1 Per Million Tokens

Ola might be far behind in the AI model race, but it’s managed to provide a hefty discount on Chinese AI model DeepSeek.

Ola Krutrim has made DeepSeek’s R1-671B model available on its platform. It’s now offering this model at an introductory price of Rs. 1 per million tokens. In comparison, DeepSeek charges Rs. 12 per million tokens.

“While we in India should be cautious with the DeepSeek app, we can totally make use of the open source model namesake, if securely deployed on Indian servers, to leapfrog our own AI progress,” wrote Ola Krutrim founder Bhavish Aggarwal on X. “Krutrim has deployed DeepSeek-R1 671B on H100s – first time anywhere in the world. It’s the most powerful open-source model available and to enable all Indian developers to access it, we will price it at ₹1/million tokens for February,” he added.

Ola Krutrim had earlier made smaller versions of DeepSeek’s models available on its platform. These included the DeepSeek-R1-Distill-Llama-70B, DeepSeek-R1-Distill-Llama-8B, DeepSeek-R1-Distill-Qwen-14B, DeepSeek-R1-Distill-Qwen-32B and the DeepSeek R1. The prices for these models ranged from Rs. 40 to Rs. 60 to generate 1 million tokens. Krutrim has now made DeepSeek’s biggest and most effective model, the R1-670, available on its platform.

Krutrim had also released details of its own models it had been working on over the last year. The results had been underwhelming — Krutrim’s models had been comparable only to small models distilled from Mistral in July last year, and were nowhere close to the cutting edge US and Chinese labs. Krutrim, however, had said it would put in Rs. 2,000 crore into developing better models this year, and a further Rs. 10,000 crore by next year.

But Ola deploying DeekSeek’s open-source model does have its own benefits. It can ensure that developers who use the model don’t need to send their data to the Chinese company, and can send it to an Indian company instead. Ola can also look to alter the model slightly to remove its Chinese guardrails — the model, for instance, doesn’t recognize Arunachal Pradesh as a part of India — but Krutrim, like Perplexity, can bypass these guardrails. India might be a long way behind in developing models, but by hosting open-source models — and giving away generous discounts — it acn hope to at least play some of a part in the ongoing AI race.

Posted in AI