DeepSeek-R1-MXFP4 is a quantized version of the DeepSeek-R1 reasoning model, optimized for AMD hardware. It supports efficient inference with reduced memory requirements while maintaining strong performance on reasoning, coding, and general language tasks. Hosted on Hugging Face, it provides model weights, chat templates, and integration instructions for developers building local AI applications.
DeepSeek R1 is a Quantised & converted weights project. It focuses on running large reasoning language models efficiently on AMD GPUs with reduced memory usage. DeepSeek R1 is an open-source project aimed at developers. DeepSeek R1 is open source under the Open Source license. It runs on the web and API.
Behind DeepSeek R1 is AMD, and it first shipped in 2025. Among its 4 catalogued features are quantized weights, AMD optimized, and chat template.
Summary written by a language model from the project’s public pages.
What PulseGate has recorded for this listing
Closest matches by what these projects do