That design was trained in part utilizing their unreleased R1 "reasoning" model. Today they have actually launched R1 itself, in addition to a whole household of new models obtained from that base.

There's a lot of things in the new release.

DeepSeek-R1-Zero seems the base model. It's over 650GB in size and, like most of their other releases, is under a tidy MIT license. DeepSeek alert that "DeepSeek-R1-Zero encounters obstacles such as limitless repetition, bad readability, and language mixing." ... so they also released:

DeepSeek-R1-which "includes cold-start information before RL" and "attains efficiency similar to OpenAI-o1 throughout mathematics, code, and reasoning jobs". That one is also MIT licensed, and is a comparable size.

I do not have the ability to run designs bigger than about 50GB (I have an M2 with 64GB of RAM), so neither of these 2 models are something I can easily have fun with myself. That's where the new distilled models are available in.

To support the research community, [forum.batman.gainedge.org](https://forum.batman.gainedge.org/index.php?action=profile