Skip to content

Free LLM API errors · Groq · checked 2026-11-13

Groq 413 Request too large (TPM): the fix

HTTP 413 · Request too large for model `<model>` … on tokens per minute (TPM): Limit 8000, Requested <n>

A single request is bigger than the model's tokens-per-minute (TPM) limit on your plan. On the free plan that is 8,000 tokens for the popular models, so no amount of waiting lets this request through.

Our own router has received this error from Groq in production.

Why it happens

  • Groq's limits are per model, and the published base limit for models like qwen/qwen3.8-27b and openai/gpt-oss-120b is 8,000 tokens per minute.
  • Prompt plus history plus the room reserved for the answer all count. Coding agents that send files or long histories pass 8,000 easily.
  • It can happen on your first request of the day: it is about size, not frequency.

How to fix it

  1. Count tokens before sending and trim: drop old turns, summarise history, send fewer files.
  2. Check Settings → Limits in the Groq console: your organisation's real limits per model are there, and another model may allow more.
  3. Send long requests to a provider with a larger per-request budget and keep Groq for short, fast ones.
  4. A paid Groq plan raises the limits.

Questions

Will retrying fix Groq's 413 Request too large?
No. The request is bigger than the per-minute token limit itself, so it fails every time until it is smaller or the limit is higher.
What is the token limit of Groq's free plan?
It is per model. The published base limit for qwen/qwen3.8-27b, openai/gpt-oss-120b and openai/gpt-oss-20b is 8,000 tokens per minute; your console shows yours.

Sources

Read on 2026-11-13. Providers change limits and messages often; the linked pages are what applies.

Limits and fine print of Groq's free tier: Groq free API. Tired of handling these by hand? ZeroLimitAI is one OpenAI-compatible endpoint that moves to the next free model when one of these comes back, and can run on your own Groq or OpenRouter key.

Other Groq errors

Stop paying per token.

One endpoint. $0 inference.

Free 7-day API key in a minute · no card. Then $99 once on your own key, or $99 a year.