Fix a ChatGPT "too many requests" error.

A "too many requests" or HTTP 429 error from OpenAI means you've hit one of three rate limits: requests per minute (RPM), tokens per minute (TPM), or tokens per day (TPD). The error is account-specific — different tiers and models have different ceilings — and the limit usually resets within a minute. The fastest fix is to wait for the window to reset, then either reduce request size, upgrade your usage tier, or switch providers in Reduz until OpenAI is willing to accept your next request. This page covers the real fixes plus the diagnostic checks for figuring out which limit you actually hit.

Check these first

• You've sent too many requests in a short window (RPM limit) — common during rapid retries or batch summarization.
• A long PDF or transcript pushed your token usage over the per-minute (TPM) cap in a single request.
• Your account tier has lower rate limits than your current workload needs — free tier is the lowest, Tier 1-5 progressively higher.
• Your OpenAI account billing is depleted or expired — the same 429 code can mean "out of credits" depending on the response body.
• A specific model has lower limits than others on your tier (GPT-5.5 may have stricter RPM than GPT-5.4 Mini).

Fix it in this order

1. 1

   Wait for the rate-limit window to reset

   Check the Retry-After header in the error response if you can see it. RPM and TPM limits reset every 60 seconds. Wait that long before retrying. Immediate retries return the same 429 and can extend the cooldown.

2. 2

   Check your usage and tier in OpenAI dashboard

   Open platform.openai.com/account/limits. Verify your current tier, the RPM/TPM/TPD limits for the model you're using, and your current usage. If you're consistently near the cap, you need to either upgrade tier or reduce volume.

3. 3

   Reduce the source size

   PDFs and long transcripts can consume 50,000+ tokens in a single request — enough to hit TPM limits in one shot. Use selected-text mode on the relevant section, summarize a page range from a long PDF, or break a 90-minute video summary into two shorter passes.

4. 4

   Try a different OpenAI model

   Each model has its own rate limit. If GPT-5.5 is throttled, GPT-5.4 Mini or GPT-4.1 may still have headroom on your tier. In Reduz settings under OpenAI, switch the active model and retry the same source.

5. 5

   Request a rate-limit increase

   For sustained high-volume use, request a tier upgrade via platform.openai.com/settings/organization/limits. Tier increases unlock with usage and time, or you can submit a manual request for production workloads.

6. 6

   Switch to another provider in Reduz

   If the 429 keeps blocking urgent work, switch Reduz to Anthropic Claude, Google Gemini, DeepSeek, or xAI Grok. The same source runs through a different model with different rate limits, and your workflow keeps moving.

Diagnosis

Rate limit (RPM/TPM) vs quota (out of credit)

The 429 code can mean either. Check the error body: "Rate limit reached" is recoverable by waiting; "You exceeded your current quota" means your account billing needs attention.

Token-driven

Long PDFs, transcripts, and big webpages can push a single request over TPM limits even when RPM is fine. If you only hit 429s on long sources, request size is the cause — not retry frequency.

Account-specific

Your friend's OpenAI account may not hit 429s on the same workload. Limits depend on your tier, your project, and the specific model. Don't assume your setup will behave the same as someone else's.

Model-specific

GPT-5.5 has stricter limits than GPT-5.4 Mini on most tiers. Switching to a smaller variant often gets the request through immediately.

Use provider choice as a pressure valve

Reduz can't increase your OpenAI rate limit — that's on OpenAI. What Reduz does is keep your Chrome summarization workflow independent from any single provider's rate limit. When OpenAI returns a 429, switch to Anthropic Claude, Google Gemini, DeepSeek, or xAI Grok from extension settings with one click. The source stays where it is; only the generation provider changes. bring your own AI key uses your own provider key, so you control the rate limit (and any tier upgrade you've made applies). Hosted Free routes through Reduz's own relay with 100 monthly credits — completely separate from your OpenAI account limits.

Frequently asked questions