udgamlabs/Gargi

Text Generation • 0.4B • Updated about 10 hours ago • 286 • 1

Note Start here. The chat model: thinking on or off, follow-up questions, documents up to ~16K tokens. Base model within 2 points of Qwen2.5-0.5B on ARC and WinoGrande with ~1,000x less training data.

udgamlabs/Gargi-GGUF

0.4B • Updated about 10 hours ago • 70

Note Q8_0 (410 MB, recommended), Q4_K_M and f16 for Ollama and llama.cpp: ollama run hf.co/udgamlabs/Gargi-GGUF:Q8_0

udgamlabs/Gargi-MLX

Text Generation • 0.4B • Updated about 10 hours ago • 244

Note bf16 build for Apple silicon (mlx-lm).

udgamlabs/Gargi-MLX-4bit

Text Generation • 0.4B • Updated about 10 hours ago • 11

Note 4-bit build for Apple silicon (mlx-lm).

udgamlabs/Gargi-Base

Text Generation • 0.4B • Updated about 10 hours ago • 79

Note The pretrained base model (32K context), for research and fine-tuning.