Critic says OpenAI hasn't fixed its model and cost problems
The author says Astra fell from Fable level to below Opus 5.5 in two weeks and is too expensive to use on the $200 plan, warning that Codex's gains could be erased by rivals' upcoming models.
I don't think OpenAI has solved their problems from the last 12 months; just two weeks ago, Astra was a Fable level model. Now it is below Opus 5.5; not just that, no one can use ASTRA on the $200 subscription because it is so expensive. Still, you actually can do some work with Opus 5.5 on the $200 plan, not to mention that Fable 5.5 is ready, and Sonnet 5.5 and Haiku are both coming, and for sure they will be better than Sol and Luna. All the work Tibo did with Codex might be wiped out in the next few weeks because they don't have strong models.
This post quotes another post: “Rumor: Claude Opus 5.5 performance caught OpenAI off guard”