Indian AI startup Sarvam has introduced its flagship large language model, Sarvam-M—a 24-billion-parameter hybrid open-weights model built on Mistral Small. Positioned as a capable, culturally relevant contender in the global LLM landscape, Sarvam-M is receiving attention for its strong performance in Indian languages, mathematics, and programming. However, some in the tech community remain cautiously optimistic, pointing out limitations in certain areas.
What does “24 billion parameters” mean?
In simple terms, parameters are like the model’s internal settings—they help it learn how to understand and generate text. The more parameters, the more detailed and sophisticated the model’s comprehension and responses can be. With 24 billion parameters, Sarvam-M sits in the mid-to-large LLM range—significantly larger than open models like Mistral 7B, but smaller than cutting-edge giants like OpenAI’s GPT-4 or Google’s Gemini 1.5 Pro. Of course, model performance also depends on data quality and training—not just size.
Where does Sarvam-M stand?
Sarvam-M may not be the biggest, but it shows impressive capabilities in domain-specific tasks, particularly maths and Indian language reasoning. It lags slightly—by about 1%—on English-dominated benchmarks like MMLU, suggesting that there’s still room to grow in global, English-heavy use cases.
How was Sarvam-M built?
Sarvam-M was trained in three key phases:
- Supervised Fine-Tuning (SFT): High-quality prompts and responses helped shape the model’s conversational skills and reduce cultural bias.
- Reinforcement Learning with Verifiable Rewards (RLVR): The model was trained to follow instructions and solve logic-heavy problems using custom-designed reward systems.
- Inference Optimization: Techniques like FP8 quantisation and advanced decoding were used to enhance efficiency and speed, though challenges remain in handling high concurrency.
Why it matters
Sarvam-M supports 10 Indian languages and performs well on competitive exam questions in Hindi. In a test combining math with romanised Indian languages, it showed an 86% improvement, underscoring its multilingual reasoning abilities.
Despite some scepticism about its global competitiveness, Sarvam-M marks a major milestone for India’s AI ecosystem. It is now openly available via Sarvam’s API and Hugging Face, inviting developers to explore, contribute, and innovate.
While not yet in the same league as the most advanced proprietary models, Sarvam-M is a significant leap forward in making AI more accessible and locally relevant, especially for communities that rely on languages other than English.
Also read: Viksit Workforce for a Viksit Bharat
Do Follow: The Mainstream formerly known as CIO News LinkedIn Account | The Mainstream formerly known as CIO News Facebook | The Mainstream formerly known as CIO News Youtube | The Mainstream formerly known as CIO News Twitter |The Mainstream formerly known as CIO News Whatsapp Channel | The Mainstream formerly known as CIO News Instagram
About us:
The Mainstream formerly known as CIO News is a premier platform dedicated to delivering latest news, updates, and insights from the tech industry. With its strong foundation of intellectual property and thought leadership, the platform is well-positioned to stay ahead of the curve and lead conversations about how technology shapes our world. From its early days as CIO News to its rebranding as The Mainstream on November 28, 2024, it has been expanding its global reach, targeting key markets in the Middle East & Africa, ASEAN, the USA, and the UK. The Mainstream is a vision to put technology at the center of every conversation, inspiring professionals and organizations to embrace the future of tech.
