Why is it in the news?
Sarvam AI’s Sarvam Vision and Bulbul V3 reportedly outperformed global AI models on India-specific benchmarks.
Marks a significant step toward building a Sovereign AI ecosystem under the ₹10,300-crore IndiaAI Mission.
Category: Science and Technology
Important Points:
● Sarvam Vision is a 3 billion-parameter vision-language model
● Performs image captioning, scene text recognition, chart interpretation, and complex table parsing
● Focuses on digitising Indian manuscripts, financial tables, and historical texts
● Trained on datasets covering all 22 official Indian languages
● Scored 84.3% on olmOCR-Bench, outperforming Gemini 3 Pro and DeepSeek OCR v2
● Achieved 93.28% accuracy on OmniDocBench v1.5
● Bulbul V3 is a text-to-speech AI model supporting 35+ professional voices across 11 Indian languages
● Optimised for Indian accents, prosody, code-switching and regional variations
● IndiaAI Mission outlay ₹10,300 crore for indigenous LLMs and AI infrastructure
● Sovereign AI ensures strategic autonomy using domestic infrastructure, data, workforce and regulatory frameworks