127 likesNew · 2d old
1-bit 763B coding model shrinks to 72 GB for local use
What is DeepSeek-V4.1-Flash-1017-GSQ-RCO-Coder-GGUF?
The DeepSeek-V4.1-Flash-1017 Coder compresses a 763-billion-parameter mixture-of-experts model to a 72-GB GGUF file by pruning half the experts and quantizing the rest to roughly 1 bit per weight. It is tuned for code generation and agentic workflows, enabling inference on a single consumer-grade GPU.
Who it is for
Developers who need a powerful code-generation model that fits on modest hardware can run it locally without a cloud API.