# sensenova/SenseNova-Loop-9B-Preview · Hugging Face — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/8261e71e26c3e20c

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-11-14 10:02 ingested 2026-11-15 02:36

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1xvysss/sensenovasensenovaloop9bpreview_hugging_face/)

## retained text



```
Very interesting model/architecture. Seems to match Qwen 3.5 27b overall at only 9b parameters. Could be useful for some and it looks like it’s not even fully cooked yet. I haven’t tried it yet. I’m not sure if it’d be slower than qwen, it’s smaller but inference parameters are \~15b and it also has mtp so idk. If you try it share your experience.
```



document fetched 2026-11-15 00:36 · canonical https://reddit.com/r/LocalLLaMA/comments/1xvysss/sensenovasensenovaloop9bpreview_hugging_face
