GPT-6 Astra draws complaints over quota drain and 503 errors
A user lists Astra's rough edges in Codex and ChatGPT Work, including fast quota consumption, capacity errors, and safety checks interrupting legitimate coding work.
This is the darker side of GPT-6 Astra that I don’t see enough people talking about. The model itself is incredible. I genuinely think it’s one of the most impressive things OpenAI has shipped. But after watching people use it heavily in Codex and ChatGPT Work, there are some problems around the experience that are getting hard to ignore. → absurd quota consumption, even for Pro 20x users → some complex tasks wiping out huge chunks of weekly usage → “model at capacity,” 503s and broken sessions → safety checks occasionally interrupting legitimate coding work → reports of citation/source failures → creative writing that can feel more restricted than competing models → some agentic workflows wasting enormous amounts of context just checking on subagents None of this makes Astra a bad model. Quite the opposite. That’s probably why it bothers me so much. You can see how insanely capable it is, and then you immediately run into all these little walls around it. I love Astra. I just don’t think the experience around Astra is anywhere near as polished as the model deserves yet. But the good news? OpenAI is clearly aware of many of these problems and is actively working on them. And with new variants coming, I’m pretty optimistic that the experience around Astra will get much better too, especially for Plus users.