Can Sarvam.AI Deliver on Its $53 Million Promise and India’s AI Mission Mandate?

Can Sarvam Justify Its $53 Million Backing and Government Mandate in India’s AI Mission?

Sarvam.AI stands as a key player in India’s growing artificial intelligence ecosystem. Recently, it attracted attention not for breakthroughs but for sparking debate.

The Bengaluru-based startup raised $53 million in Series A funding from major investors like Lightspeed and Peak XV Partners. It is one of the most well-funded deep tech startups in India and a flagship participant in the Centre’s IndiaAI Mission aimed at building the country’s first sovereign foundational AI model.

The Launch of Sarvam-M: Expectations vs Reality

Sarvam recently launched Sarvam-M, a 24-billion-parameter open-weight large language model (LLM) based on Mistral Small, an open-source LLM by French startup Mistral AI. This marks the first in a planned series of releases showcasing their research.

The company highlights Sarvam-M’s multilingual support for Indian languages along with programming and mathematical reasoning capabilities. However, the tech community's response has been mixed. While some applaud the initiative, many question its originality and significance.

Critics point out that Sarvam has essentially fine-tuned an existing model rather than building a new foundational model. Puneet Pandey, Founder of Ojas Softech Pvt. Ltd., states, “Fine-tuning is now a commodity. There are hundreds of Indian companies, including ours, that have done it. So, what’s new here?” He adds that benchmarks show only marginal improvements over older models like LlaMA and Claude, which have lost relevance.

Pressure and Performance

Given Sarvam’s strong funding and government links, many expected more significant innovation. Industry insiders speculate that the company may be under pressure to demonstrate progress publicly as part of the IndiaAI Mission.

Despite this, Sarvam-M’s initial downloads have been underwhelming. This raises concerns about the model’s adoption and practical demand.

The Need for Indian Language AI

Experts acknowledge a genuine demand for AI models that understand Indian languages natively. Pandey notes, “Existing global models like OpenAI’s GPT or Google’s Gemini dedicate only 1-3% of their training data to Indian languages. They mostly translate from English rather than comprehend regional languages directly.”

Support and Critique from Industry Leaders

Sridhar Vembu, Founder of Zoho, supports Sarvam’s efforts, emphasizing that immediate success is not the only measure of long-term impact. Jaspreet Bindra, Founder of AI & Beyond, offers a balanced perspective. He says, “Building a 24-billion-parameter multilingual model supporting 10 Indian languages, especially in a developing deep tech environment, is commendable.”

Bindra highlights that Sarvam-M performs well on India-specific benchmarks such as answering competitive exam questions in Hindi accurately. He suggests that limited adoption might be due to insufficient tooling, poor documentation, or a lack of urgent India-specific use cases.

The Way Forward: Focusing on Real-World Indian Use Cases

To increase adoption, Bindra recommends focusing on practical applications in agriculture, rural governance, and healthcare. These sectors could benefit significantly from AI solutions tailored to local languages and conditions.

For Sarvam.AI, the challenge is clear: move beyond symbolic product launches and develop solutions that genuinely meet the needs of India’s diverse population.

• Prove that government-backed investments translate into impactful technology
• Build tools and documentation that encourage developer and industry use
• Focus on real-world problems with measurable benefits for Indian users

As India’s AI ecosystem matures, the expectations for sovereign models will only rise. Sarvam’s next steps will determine if it leads or simply participates in the country’s AI efforts.

For those in government roles interested in AI developments and opportunities, understanding these dynamics is crucial. To explore relevant AI courses and training, visit Complete AI Training.