Valeciela

/

KansenSakura-Symbiosis-12B

like 6

No. The root model for this merge and all the models involved in the merge is Mistral Nemo 2407, which is not a vision model.

Okay, that's a shame. Thanks for answering.

No. The root model for this merge and all the models involved in the merge is Mistral Nemo 2407, which is not a vision model.

I found and tried this model: https://huggingface.co/bartowski/mistral-community_pixtral-12b-GGUF/blob/main/mmproj-mistral-community_pixtral-12b-f16.gguf

It describes the shapes almost correctly, but the colors are glitchy, where human skin is purple.

Is there a similar model with vision?

If what you care about is vision accuracy and are restricted to a size range around 12B, then I would try one of the Qwen3 VL models, such as this.