Comment by sauwan

I'd be surprised if this was a new base model. It sounds like they just did some post-training RL tuning to make this version specifically stronger for coding, at the expense of other priorities.