AI Model
Published Aug 24, 2026
Sarvam 2B
Efficient 2-billion parameter Indic language model optimized for low-latency voice and text compute.
Profile & Intelligence Breakdown
Sarvam 2B is an open-source 2-billion parameter foundational language model built by Sarvam AI. Specifically architecture-tuned for Indic language performance, Sarvam 2B outperforms larger models on Hindi, Bengali, Gujarati, and Tamil benchmarks while running efficiently on edge compute hardware.
Historical Milestones
Chronology & Major Events
Aug 15, 2024
Model Weights Release
Sarvam AI released Sarvam-2B weights on HuggingFace for open research.
Official Citations & Sources
Technical Specifications
- parameters
- 2B
- context length
- 8k tokens
- license
- Open Source
- developer
- Sarvam AI