Dflash support for gemma4

[Ofir408]

Hi,
I was looking through the repo, and it seems the current supported models are mainly Qwen3/Qwen3.5, gpt-oss, and Llama 3.1, with no Gemma-family support listed yet.
Is it possible to add support for gemma4 models?
Thank you!

[jianc99]

We are working on Gemma 4, will support multiple model sizes.

[trivedip]

Ty, hope we will also get MOE and Dense support! this looks really neat.

[itzpingcat]

PLS

[Naist4869]

666

[pdramos1]

What is the timeline on how long this takes to get done? Just trying to coordinate my expectation and when to look for the release.

[jianc99]

@pdramos1 We’ve completed the training data regeneration for Gemma 4 26B and 31B. I can’t guarantee an exact timeline, but we’ll do our best to train the draft models and release them as soon as possible!