Our 2-bit GGUF format: Qwen3-4B and Llama-3.1-8B, 85% smaller, no retraining. Needs our patched llama.cpp (CUDA).
-
Pingala-8 2-bit LLMs
📉2-bit GGUF LLMs for llama.cpp, 85% smaller, no retraining
-
Genrius-AI/Qwen3-4B-Pingala8-2bit
Text Generation • 1B • Updated • 89 -
Genrius-AI/Llama-3.1-8B-Pingala8-2bit
Text Generation • 3B • Updated • 124 • 1 -
Genrius-AI/pingala8-benchmarks
Viewer • Updated • 30