# Best Omni Model under 40B parameters currently — retained record

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/c8cd1703e894550d

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


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-09-25 11:47 ingested 2026-09-26 02:38

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1wpu6wf/best_omni_model_under_40b_parameters_currently/)

## retained text



```
I am searching for a fully omni modal ie Voice recognition and speech generation like qwen 3 omni 30b a3b is their any newer model or finetune which is more capable or efficient in this category? gemma 4 is great but it does not support the speech generation.
```



document fetched 2026-09-26 00:41 · canonical https://reddit.com/r/LocalLLaMA/comments/1wpu6wf/best_omni_model_under_40b_parameters_currently
