:
Eight months after its launch, the initial euphoria surrounding DeepSeek-R1 is cooling, giving way to a more pragmatic approach regarding its usage in India. While some developers find it a valuable tool, application builders and end-users are carefully weighing their options, often preferring solutions that better align with their immediate needs for cost-effectiveness and ease of deployment.
For model developers, DeepSeek-R1 was initially seen as a godsend. Prasenjit Dey, an Assistant Professor of Computer Science and Engineering at NIT Rourkela, points out that its primary strength lies in providing access to “very large pretrained models without heavy GPU infrastructure.”
“This capability has significantly reduced our prototyping cycles from weeks to days and lowered the entry barrier for students, making developers enthusiastic about DeepSeek-R1 for fast prototyping with large models.
Dey notes that it serves as a complementary platform that accelerates early-stage research rather than a complete replacement for traditional infrastructure.
However, for use-case builders and smaller organisations, the landscape appears quite different. Rakesh Dubbudu, Founder of Tagore AI, suggests that “there is no additional advantage with open source” for entities that are not very large and have specific use cases.
He explains that API prices have fallen drastically over the last two to three years, making them “comparable” or even more cost-effective than open-source options in some cases.
He said that maintaining your own infrastructure is a “burden that requires specialised skills and a large DevOps team, making it not a practical idea for many users”.
“Application builders rarely buy GPUs and prefer plug-and-play solutions that allow them to start working immediately rather than setting up infrastructure. API prices have fallen so significantly that there’s no reason for anybody to choose other options,” Dubbudu pointed out.
DeepSeek-R1 also faces significant market realities and stiff competition in India. AI industry analyst Kashyap Kompella notes that despite generating substantial curiosity and app downloads, DeepSeek’s growth was off a small base.
In India, DeepSeek had around 3.7 million monthly active users, while ChatGPT, despite being banned in China and Russia, commanded 71 million as of April 2025. Kompella stresses that DeepSeek faces significant competition in India, with OpenAI launching India-specific pricing for GPT-5, Google offering free Gemini Pro for college students, and Perplexity AI partnering with Airtel.
“Meta has also had initiatives to grow the ecosystem around Llama, and other Chinese open-weight models like Alibaba’s Qwen maintain traction among Indian developers for local deployment,” he observed.
Ankush Sabharwal, Founder and CEO of CoRover and BharatGPT, highlights a major drawback for the Indian market: DeepSeek-R1 doesn’t support Indian languages, which he considers a major drawback for users in India.
While acknowledging DeepSeek-R1’s exceptional performance in mathematics, programming, and general logic, along with its efficiency and affordability, Sabharwal points out that BharatGPT supports over 14 Indian languages.
Kiran Babu, Founder and CEO of content marketing solutions startup Rava.ai said that they used it as a draft engine—fast, structured, and cost-effective—within their multi-model router, while in-built guardrails and editorial layers ensured accuracy, compliance, and brand safety.
“It offers strong logical structuring but higher hallucination rates than closed models. It offers about 40% lower cost per asset with comparable editing effort to GPT-based pipelines,” he said.
He, however, noted that some clients are cautious when they hear that DeepSeek is a Chinese model, raising concerns about security and compliance.
“We tell them that all data is encrypted, never used for training, and that DeepSeek is only one part of our multi-model router, with fallback to non-Chinese providers,” he said.
Published on August 20, 2025