Groq runs open models on its own hardware and is known for answering very quickly. Its free plan is enough for everyday questions, and getting a key takes about a minute.
First, a common mix-up: Groq (with a q) is an inference company. Grok (with a k) is xAI’s chatbot. They are unrelated, and a key from one does not work with the other.
Create the key
- Open the API Keys page in the Groq console and sign in or create an account.
- Click Create API Key and give it a name you will recognize later.
- Copy the key straight away and store it somewhere private. You will not be able to see it again; if you lose it, create a new one.
Free plan rate limits
Groq sets limits per model. On the free plan, the published limits for the models most useful for questions are:
| Model | Requests per minute | Requests per day | Tokens per minute | Tokens per day |
|---|---|---|---|---|
| GPT-OSS 120B | 30 | 1,000 | 8,000 | 200,000 |
| GPT-OSS 20B | 30 | 1,000 | 8,000 | 200,000 |
| Qwen 3.8 27B (accepts images) | 30 | 1,000 | 8,000 | 200,000 |
Three details matter in practice:
- Limits apply to your whole organization, not to each key. A second key shares the same allowance.
- The daily token cap usually runs out before the request cap. A question with a long answer can use a few thousand tokens, so a heavy day ends well before 1,000 requests.
- When you go over, the API answers with
429 Too Many Requestsand aretry-afterheader that says how many seconds to wait.
The numbers for your own account are under Settings → Limits in the console. Treat that page as the truth; published tables change.
Which models are not free
Not every model in Groq’s catalog is on the free plan. Llama 3.1 8B and Llama 3.3 70B are now listed for Enterprise use, and so are newer preview models such as MiniMax M2.7. If a model is missing from the free limits table, assume it is not free for your account.
Which model to pick
- GPT-OSS 120B for text questions. It reasons well and is still fast.
- GPT-OSS 20B when you want the quickest possible answer to a simple question.
- Qwen 3.8 27B when the question is an image: a chart, a diagram or a formula. It is the only one of the three that accepts images (up to three per request, and each image counts as about 2,048 input tokens). Groq lists it as a preview model, so it can change at short notice.
Keep the key safe
Do not paste the key into screenshots, chat messages or a repository. If you write your own scripts, read it from an environment variable (GROQ_API_KEY) instead of typing it into the code. Keys that end up in public code are found and abused quickly; if that happens, delete the key in the console and create a new one.
Common questions
Is the Groq API free? Yes, for the models in the free limits table above, with the daily caps shown.
What happens when I reach the limit? Requests fail with a 429 error until the minute or the day resets.
Does Groq work for questions with images? Yes, with Qwen 3.8 27B. The two GPT-OSS models read text only.
Using the key in LuciQA
LuciQA is a free desktop app that answers questions about what is on your screen. Open Settings, pick Groq, paste the key and save. For a text question LuciQA uses the model you chose; when you send an image it switches to Qwen for that request, so you do not have to change the model yourself.
See the Groq setup guide for the same steps, or the free AI APIs comparison for the other providers with a free tier.