# Too good to be true? ISTA's GSQ-RCO 2-bit DeepSeek-V4.1-Flash scoring the same as the base model — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/d748d658b4cb66e8

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-10-25 03:48 ingested 2026-10-26 02:40

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1xes5x0/too_good_to_be_true_istas_gsqrco_2bit/)

## retained text



```
Has anyone tried ISTA's GSQ-RCO GGUFs for DeepSeek-V4.1-Flash? Does the quality feel as

good as their benchmarks show (IQ2\_XXS matching the base)?

Their benchmarks are saturated (90%+, and AIME25 is 100 for every version), so they

probably can't show the degradation, and I couldn't find any KLD for GSQ-RCO models.

Has anyone measured KLD or PPL? What KLD would you expect at \~2 bpw?

[https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF](https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-GGUF)

(English isn't my first language, I used an LLM to help translate this post.)
```



document fetched 2026-10-26 00:38 · canonical https://reddit.com/r/LocalLLaMA/comments/1xes5x0/too_good_to_be_true_istas_gsqrco_2bit
