← sources

ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO

source: reddit (reddit:LocalLLaMA)published 2026-10-23 23:37ingested 2026-10-24 02:40

retained text

A brand-new release from ISTA-DASLab with their GSQ-RCO smart quantization approach.

Now DeepSeek-V4.1-Flash-1017

The quants are available on Hugginface:
https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF

Llama.cpp is not supported because the original architecture support is not yet merged:

https://github.com/ggml-org/llama.cpp/pull/28696

https://github.com/ggml-org/llama.cpp/pull/30817

Strata support is mentioned in the model card but I'm pretty sure the original Strata only supports Qwen3.8-Flash-Next so most likely what they meant was one of the forks based on Strata. Most likely [Xolotl](https://github.com/Indras-Mirror/Xolotl) or [Maya](https://github.com/mw00/project-maya)

document fetched 2026-10-24 10:42 · canonical https://reddit.com/r/LocalLLaMA/comments/1xdtk7m/istadaslabdeepseekv41flash1017gsqrco