DeepSeek Flash GSQ-RCO GGUF
ISTA-DASLab's mixed-precision GGUF builds of DeepSeek V4.1 Flash 1017, with a documented Strata runtime path and substantial disk requirements.
What happened lately
DeepSeek Flash GSQ-RCO GGUF timeline
ISTA-DASLab publishes mixed-precision GGUF builds of DeepSeek V4.1 Flash 1017 using GSQ and RCO. The model repository was created October 23, 2026 and appeared in AITrending’s public Hugging Face trending discovery on October 27. That observation is not a new release date or a measurement of growth.
The English model card documents 1.75, 2.00 and 2.25 bits-per-weight backbone variants, each paired with a separate Engram file. It recommends IQ2_XXS and directs users to Strata. Its dated compatibility warning says upstream llama.cpp did not support this model as of October 23; later example commands in the card are conditional on runtime support.
Strata’s English README documents NVIDIA RTX 30, 40 and 50 series support on Windows and Linux for this model, plus Radeon AI PRO R9700 on Linux. The documented 12 GB NVIDIA GPU configurations also use substantial system RAM and SSD storage. The runtime is available through desktop installers or a terminal launcher, with no account required according to its authors. AITrending has not reproduced the hardware performance claims.