GOOD NEWS!!! 🎉
OAI fixed the bug that randomly routed Pro / Thinking to GPT-5.5 mini.
They say ~3% of turns were affected; my own testing hit 22%.
So yeah, I wasn’t imagining it — the fucking router was actually broken 😂
I’ve already reached out to the OpenAI Support team and reported an issue where GPT-5.6 Sol on mobile is sometimes automatically routed to GPT-5.5 mini.
We found that ChatGPT5.6luna is absolutely perfect for data cleaning - it's like discovering a new continent. The quota is practically unlimited, even though I'm just a Plus user
ChatGPT Android model routing seems weird:
Thinking OFF → GPT-5.6 Sol
Thinking LOW/HIGH → GPT-5.5-mini
Reproducible on my Plus account. Why does enabling Thinking apparently downgrade the model?
Bug or intentional routing?
@ChatGPT@OpenAI
Introducing the next-generation rreasoning architecture that hasn't been fully developed yet. Thanks to @thsottiaux for providing the reset. We're developing it using ChatGPT5.6sol max-Fast - feel free to handle this message however you want