What you see
API Error (529 {"type":"error","error":{"type":"overloaded_error","message":"Overloaded"}}) · Retrying in 1 seconds… (attempt 1/10)
API Error (529 {"type":"error","error":{"type":"overloaded_error","message":"Overloaded"}}) · Retrying in 2 seconds… (attempt 3/10)Those lines are from #3572, the July 2025 wave that collected 274 comments — many from people on the $200 Max plan, which is the point: 529 is capacity, and a bigger subscription does not reserve any. The error is still being reported now: #91817 ("excessive retry attempts"), #91849 and #91825 were all filed on September 3, 2026.
Why it happens
- Demand spikes on one model. New releases and weekday US mornings are the classic peaks; the newest flagship is the most contended.
- Your own fan-out. #90043 scanned 60 days of transcripts and found most 529s arrived in bursts: an orchestrator launches N agents together, they hit the API together, and most of the batch fails in the same minute.
- Lost work. #89267 and a related report note that a 529 can terminate a background subagent instead of retrying it — so a burst doesn't just slow you down, it drops tasks.
Fixes
- 1. Let the built-in retry run. Most 529s clear within the 10 attempts.
- 2. Switch tier:
/modelto Sonnet (or Opus) for the current step; come back later. - 3. Run fewer subagents at once and give fan-outs a stagger; a burst of simultaneous first requests is what turns one 529 into ten.
- 4. Re-run dropped subagents explicitly — check that every unit of a fan-out actually finished.
- 5. Off-peak for the big jobs: long autonomous runs are far less likely to hit capacity outside US working hours.
529 vs 500 vs 429
- 529 Overloaded — capacity, everyone, temporary. Wait or switch model.
- 500 — server failure. See API Error 500.
- 429 — your own limit (requests, tokens, concurrency). See rate limits.
On this gateway
The gateway spreads traffic across several upstream accounts and retries a 529 on another before giving up, so a single saturated account doesn't stop your session. When a model is saturated everywhere, you'll still see it — pick another tier with /model. Same setup as always: ANTHROPIC_BASE_URL=https://aiprimetech.io and a gateway key.
Frequently asked questions
What does the Claude overloaded error mean?
The model has no spare capacity at that moment (HTTP 529, overloaded_error). It affects everyone using that model and clears on its own.
How do I fix Claude API Error 529?
Let Claude Code retry, switch model tier with /model, and run fewer parallel subagents. For large jobs, pick off-peak hours.
Does a Max subscription avoid overloaded errors?
No. 529 is about capacity, not plan limits — Max users report it as often as anyone.
Run Claude Code on the gateway
Same models, two environment variables, credits at 7.69× face value — or a flat-rate unlimited plan.
Get an API key See unlimited plansAI Prime Tech is an independent API gateway and is not affiliated with, endorsed by, or sponsored by Anthropic. “Claude” and “Claude Code” are trademarks of Anthropic. Claude Code features described here follow Anthropic’s public documentation at the time of writing and change frequently; prices and model lists on this page are read from this gateway’s live settings.