[BUG] Server is temporarily limiting requests

Found 3 possible duplicate issues:

1. https://github.com/anthropics/claude-code/issues/63713
2. https://github.com/anthropics/claude-code/issues/63761
3. https://github.com/anthropics/claude-code/issues/53922

This issue will be automatically closed as a duplicate in 3 days.

• If your issue is a duplicate, please close it and 👍 the existing issue instead
• To prevent auto-closure, add a comment or 👎 this comment

🤖 Generated with Claude Code

I am facing the same issue.. Whats wrong with Claude.. I just got my 5 hour limit reset and then within next 5 min 100% again.
And I can see the same screenshot as that of attached in the bug.

its happening to me too, though i have the $20 plan and i havint used claude code for over a day and my usage limit is fresh and unused and im still getting this same error

The problem seems to be linked with the session I am in. The faulty session works fine for all models, except Sonnet (that still gives me a "pay more to receive service" error). If I open a new Session, then Sonnet works again as if nothing was wrong. In this new session I am now somewhere around 2M tokens used with Sonnet on "Medium effort" and I have not seen any more api errors.
@Manish-Kumar94 , @aycmcompany maybe that solves your situation as well

...and just as I wrote the update (literally 10 seconds later) the api error appeared again. Still I was able to work for 3 hours at least...

Server is temporarily limiting requests
Too many requests right now — try again in a moment.
View details
API Error: Usage credits required for 1M context · turn on usage credits at claude.ai/settings/usage, or use --model to switch to standard context

Same problem here, and the now-closed #62426 ("completed", but clearly unresolved given the flood of open duplicates — #64327, #64088, #63862, #63816, #63789, #63703, #63173 …) already nailed the core point.

Adding a data point: this is markedly worse on the 1M-context Opus model. It draws from a separate, much tighter limit pool, so even at top paid tier with a freshly-reset usage window, Server is temporarily limiting requests (not your usage limit) · Rate limited fires within minutes. Some of us also see the variant Usage credits required for 1M context · … use --model to switch to standard context.

The contradiction users keep raising is the actual issue: Anthropic's own docs actively promote parallel / multi-session work — Run agents in parallel, Orchestrate teams of Claude Code sessions, agent view, "scaling across parallel sessions" in Best Practices, desktop parallel sessions with Git isolation — yet the account-level shared rate-limit pool plus an infrastructure burst limiter punish exactly that pattern.

Telling users to "disable 1M context" or "switch to a smaller model" is not a fix — it means we can't actually use the models and features we pay for. A workaround that disables the thing we paid for defeats the purpose.

Asks:

1. Scale infrastructure request-rate limits meaningfully with paid tier so parallel-agent patterns are viable.
2. Do not gate 1M context (or model choice) behind a tighter throttle that effectively blocks paid usage.
3. Make the error actionable: state which limit was hit (plan usage vs infra throttle), current consumption, and reset time.
4. Resolve the gap between "we recommend parallel sessions" and "rate limits don't support them."

the same issue
requested support but that doesnt work
the issue is for 4 days already on claude desktop

I have now switched over to working with "Opus 4.8" (not the 1M context model, but the normal one). Claude desktop anyway shows me a 1M context window (and is using it). After three intense days I came to the following conclusion:

1. No api or usage errors with Opus 4.8 (high effort)
2. No difference to Sonnet - actually I like Opus 4.8 more than Sonnet for Java development
3. With about 4 hours intense development per day I won't be able to come close to my quotas (I reach like 40% weekly total) - I guess I'm using between 3M and 10M tokens per day

<img width="874" height="340" alt="Image" src="https://github.com/user-attachments/assets/bd1aee39-0cfe-4d14-8464-5a19d4a4dabd" />

Same issue here yesterday with Claude Code Sonnet 4.6 were totally down, but switching to the older Sonnet 4.5 worked perfectly.
/model claude-sonnet-4-5

Facing the same problem.

I understand the need to occasionally temporarily limit requests. But the error message should explain how to proceed. I got

API Error: Usage credits required for 1M context
turn on usage credits at claude.ai/settings/usage, or use --model to switch to standard context

My context window is at 75.3 k, nowhere near 200k and certainly not approaching 1M. And I can't see any UI option to actually select the context window size – none of the 5 selectable models have a "(1M)" appended – maybe because I'm on Pro?

<img width="460" height="246" alt="Image" src="https://github.com/user-attachments/assets/dc916b10-d877-4ef0-b4b6-c8f7571206a8" />

Plus, the message is not helpful. I'm in the desktop app, I can't use --model, and even if /model does the same, which model should I choose?