You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Can you add Qwen 3.5 models to the supported models?
Proposed solution
Idk but I'm currently researching myself how to convert the model to .pte format so it can be compatible with react-native-executorch
Alternative solutions
No response
Benefits to React Native ExecuTorch
I tested the model in Ollama locally and it's much better than anything I tested before, even the 0.8b version is so good, I believe it will be a big win to the package if you supported it
Activity
chmjkb commentedon Mar 5, 2026
Hey @markshenouda, thanks for your interest! We definitely want to add support for Qwen 3.5 and we will look into it soon cc @NorbertKlockiewicz
msluszniak commentedon Mar 23, 2026
I guess this is what we needed: pytorch/executorch#18388, cc: @NorbertKlockiewicz
mergennachin commentedon Mar 23, 2026
@msluszniak -- the PR you mentioned above is MoE model. Is that what you wanted?
The dense variant of qwen3.5 model is https://github.com/pytorch/executorch/tree/main/examples/models/qwen3_5, there's also an outstanding PR here pytorch/executorch#18088
msluszniak commentedon Mar 23, 2026
Indeed, pytorch/executorch#18088 is what we need, thank you for pointing out :))
chore: add Qwen3.5 and Bielik v3.0 models (software-mansion#1096)
msluszniak commentedon Oct 22, 2026
I think this is not true, to fully support qwen 3.5, we should iterate over this one.
msluszniak commentedon Oct 22, 2026
MiniCPM5-2B is a comparable on-device alternative.