Rumor: GPT-6 Sol is coming soon, with Astra as the base model
A post claims Astra is the base model for OpenAI's RL post-training, that GPT-6 Sol will be better and cheaper and likely uncapped for Plus users, and that Astra on low reasoning already far outperforms GPT-5.6 Sol on max.

GPT-6 Sol Is Probably Coming Very Soon 🛰️ >Astra is basically the base model — OpenAI now does RL/post-training from here >Expect GPT-6 Sol to be better, cheaper, and likely the uncapped model for Plus-tier users >BTW: Astra on "low" reasoning effort is already far better than GPT-5.6 Sol on "max" >That gap alone tells you how big this generational jump really is Sol might end up being the model most people actually use day to day.