I'm also on the Claude Team Premium plan and can always fall back to PPQdotAI (nprofile…4u96) on opencode. It's just become really frustrating to work with ChatGPT. Guess it's time to cancel that plan and bump my spend on PPQ. The Chinese models are either close or sometimes better but I dumped a lot of memory for some of my stuff on ChatGPT and this just never happened until about a week ago. I had multiple sessions in parallel on opencode using their legit headless oauth, worked on issues, PRs, CI, planned features and now I'm hitting limits from a one session conversation with 5.6 on Medium with a complexity that Grok/Deepseek free tiers do without complaints.
My suspicion is that their credits accounting has been built with AI and they don't understand how it even works anymore.
Guess I'll do without them starting next month.
