Jifeng Dai's team open-sources its first model, 36kr reports

36kr has published a short report saying that Jifeng Dai's team has open-sourced its first model, with a headline describing the work as something built with AI. The item carries little beyond that framing.

Looking at what can be checked, three things are confirmed: the actor is Jifeng Dai's team, the action is an open-source release, and the object is that team's first model. Parameter count, architecture, license type and where the weights live do not appear in the public text.

The phrase about the model being built with AI is also left unexplained. It could mean training data generated with AI, a training pipeline driven by AI, or a model that took part in its own development; the available wording does not settle which.

So this reads as a signal rather than a technical description. A Chinese research team pushing its first model straight onto an open channel says something about a distribution shift toward community releases, but whether the model is worth using cannot be judged until a model card and weight repository are visible.

Two things to watch: whether the team backfills a model card, license and evaluation data in the coming days, and whether the built-with-AI angle gets made concrete in follow-up coverage.