Sarvam Releases Open-Weight Multilingual LLMs With 30B and 105B Parameters for India

Summary

Indian AI startup Sarvam has released two open-weight multilingual large language models with 30 billion and 105 billion parameters, first unveiled at the AI Impact Summit 2026 in New Delhi. The models are available under the permissive Apache 2.0 license, targeting India's linguistically diverse market. The release reflects the growing trend of region-specific and language-specific foundation models emerging outside of the dominant US and Chinese AI labs. Open-weight multilingual models could