Free LLM APIs · checked 2026-10-22
Groq: what “free” actually gets you
Groq runs open models on its own hardware and is known for fast responses. Its free plan needs no card and gives each model its own limits, so several models together go further than one.
- Free allowance
- ~1,000 requests/day per model
- Card needed
- No
- Trains on prompts
- Not stated on its data page; no retention by default
- Main catch
- 8,000 tokens per minute per model: long prompts don't fit
- Base URL
https://api.groq.com/openai/v1
Limits
- Limits are per model. For qwen/qwen3.8-27b, openai/gpt-oss-120b and openai/gpt-oss-20b the published base limits are 30 requests per minute, 1,000 per day and 8,000 tokens per minute.
- Your account's own limits are on console.groq.com/settings/limits. That page, not the docs, is what applies to you.
The fine print
- 8,000 tokens per minute means a single request with a long prompt (a coding agent sending a whole repository, a long document) is rejected even if it's your first of the day.
- Check Settings → Billing. If your organisation has a card and is on the on-demand plan, usage is billed. The free limits in the docs are not proof you are on the free plan.
- Groq's model list changes: models move between free and paid plans. List the models your key can call (GET /models) instead of hard-coding one.
Your data
- By default Groq does not retain inference requests. Inputs and outputs may be logged for up to 30 days for reliability and abuse review.
- Admins can turn on Zero Data Retention in Data Controls.
- Groq's data page doesn't mention training on customer data either way.
Get started
- Create an account at console.groq.com (no card needed) and an API key.
- Use base URL https://api.groq.com/openai/v1 with any OpenAI-compatible client.
- Check Settings → Limits and Settings → Billing before relying on it.
Questions
- Is the Groq API free?
- There is a free plan without a card. Each model has its own limits, for example 30 requests a minute and 1,000 a day for qwen/qwen3.8-27b.
- Why does Groq reject my request with "Request too large"?
- The free plan caps tokens per minute per model (8,000 for the models above). A request bigger than that can't succeed on the free plan, however few requests you've made.
- Does Groq store my prompts?
- Not by default. They may be logged for up to 30 days for reliability and abuse review, and admins can turn on Zero Data Retention.
Sources
Read on 2026-10-22. Providers change limits and terms often; the linked pages are what applies.
Which free models are worth using today, and who serves them: state of free LLM APIs and the open ranking on GitHub. One endpoint across free models, with your own Groq or OpenRouter key if you have one: ZeroLimitAI.