Gemma 4 12B: A unified, encoder-free multimodal model

Re:Gemma 4 12B: A unified, encoder-free multimodal model by ricardobayes on Thursday June 04, 2026 @09:28AM (#48398354)

You can run it, however those low quantized models (iQ2, iQ4, Q2) will very likely underperform the 9B versions at Q6/Q8.

[ Parent | on HN ]

Re:Gemma 4 12B: A unified, encoder-free multimodal model by kanemcgrath on Thursday June 04, 2026 @05:33PM (#48404921)

Something about qwen models hold up really well even at low quants. for most other models anything under q5 is cooked, but on 35B-A3B I can get a lot of things done even at q3_xl. It is definitely better than full precision 9B

[ Parent | on HN ]