eulogik/Bharat-Tiny-LLM-v3
Text Generation • 2B • Updated • 2.65k • • 4
Next-gen Indic edge LLMs (Bharat-Tiny-LLM), PolyWhisper ASR, and the Brahmi Tokenizer with 33-45% compression for Devanagari.
Note 1.7B Hindi/Hinglish edge LLM on Qwen3 with tool-use and GSM8K math reasoning
Note Indic multilingual speech recognition (5 languages) with dual LoRA experts
Note MLX-optimized Indic LLM for Apple Silicon Macs
Note Code-switching language router for Indic speech streams
Note Instruction-tuning conversational dataset for Indic edge LLMs