# Kind of enjoying the ISTA (GSQ-RCO) vs Unsloth beef — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/1940a51dcf599d64

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-11-02 20:43 ingested 2026-11-03 02:42

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1xm8xhn/kind_of_enjoying_the_ista_gsqrco_vs_unsloth_beef/)

## retained text



```
I'm not on anyone's side, I just love the competition. Both teams pushing each other means better quants for all of us.

What I don't understand is why ISTA doesn't publish KLD for their quants. It doesn't depend on how long the model thinks, so it would make this a lot easier to compare. The only KLD I've seen for GSQ-RCO is in Unsloth's charts. Maybe there's a reason I'm missing.

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



document fetched 2026-11-03 00:43 · canonical https://reddit.com/r/LocalLLaMA/comments/1xm8xhn/kind_of_enjoying_the_ista_gsqrco_vs_unsloth_beef
