ETA for Qwen3.5-122B/397B and GLM-5.1 DFlash draft models?

[01lin]

Question

The README lists the following models as coming soon:

• Qwen3.5-122B
• Qwen3.5-397B
• GLM-5.1

Could you share an estimated timeline for when these draft models will be open-sourced?

Specifically:

1. Qwen3.5-122B / 397B – Are these actively in training? Any rough ETA?
2. GLM-5.1 – Same question — is this already trained and pending release, or still in progress?

These are highly anticipated models in the community and many deployment scenarios depend on them. Any update would be greatly appreciated!

Environment (for context)

• Deployment backend: vLLM / vLLM-ascend
• Target use case: Production inference acceleration for large MoE models

[jianc99]

We just released the DFlash draft model for Qwen3.5 122B, which is a lightweight 4 layers draft model with interleaved sliding window layers. And we are working on 397B and GLM5.1!

[my-other-github-account]

Very interested in GLM5.1 - please let us know!

[dongyibo]

looking forward to it!