DeepSeek V4.1-Flash 763B Runs Locally on 12GB GPU with Quantization
DeepSeek V4.1-Flash (763B) runs on consumer GPUs with 12 GB VRAM using GSQ‑RCO quantization, reaching up to 231 tokens/s input processing, but requires ~239 GB storage.
cpaua2 min8Filter by tag
#local-inference
All tagscpaua2 min8