50x Faster Post-Training

Re:50x Faster Post-Training by kostolansky on Friday March 13, 2026 @02:15PM (#47367716)

We took another step toward making open source models truly open, namely by creating a training stack that actually allows for finetuning of a large frontier OSS model, Kimi K2 Thinking. (The OSS stack for big models is surprisingly pretty abysmal these days!)

There is lots of value to be unlocked by people using language models for their own purposes, and our work here hopefully moves the needle towards making that more accessible to more people. (The training code will be released soon, pending safety testing.)

We are very excited by what we can all build :D

- Tim from WSL

[ Parent | on HN ]

Re:50x Faster Post-Training by addiefoote8 on Friday March 13, 2026 @02:51PM (#47368121)

I'm also excited about the research that could be enabled by having weight-level access and fine tuning access on frontier open source models. There's a lot of interesting behavior that just doesn't exist in 8B parameter models, not to mention with architectural and training differences.

[ Parent | on HN ]