Is the Groq API free? Free tier limits, models, and how to use it on iPhone and Mac

Yes, the Groq API has a free plan. As of November 2026 it covers three chat models, OpenAI’s gpt-oss-120b and gpt-oss-20b and Qwen 3.8 27B, each with up to 30 requests a minute, 1,000 requests a day, 8,000 tokens a minute and 200,000 tokens a day (Groq). You need a Groq account and an API key; Groq asks for a payment method only when you upgrade to its paid Developer plan. Because Groq’s API is OpenAI-compatible, you can use it from an iPhone, iPad or Mac in Modelbase as a custom server.

What is Groq?

Groq is an AI inference company: it runs open models on its own chips (it calls them LPUs) and sells access to them through an API, which it describes as a cloud for fast inference. It isn’t the same as Grok, the chatbot from xAI.

Groq free tier limits (November 2026)

Groq’s rate-limits page lists the free plan’s limits per model. These are the ones that matter for chat:

Model ID	Requests/min	Requests/day	Tokens/min	Tokens/day
openai/gpt-oss-120b	30	1,000	8,000	200,000
openai/gpt-oss-20b	30	1,000	8,000	200,000
qwen/qwen3.8-27b	30	1,000	8,000	200,000
whisper-large-v3 (speech to text)	20	2,000

Source: Groq rate limits, free plan, November 2026.

Some things to know about these numbers:

• They apply to your organization, not to each key or user.
• You hit whichever limit comes first. With long conversations, the 8,000 tokens a minute can run out before the request limits do: a big prompt plus a long answer can use a good part of it in one go.
• Over the limit, the API answers “429 Too Many Requests” with a retry-after header that says how many seconds to wait (Groq).

Which models are no longer free on Groq?

A lot of older guides recommend Groq’s free Llama models. Those have changed: Llama 3.1 8B Instant and Llama 3.3 70B Versatile were removed for free and Developer users in August 2026, and Qwen 3 32B, Llama 4 Scout and Kimi K2 have been shut down (Groq deprecations).

What the free Groq models can do

Model	Context	Images	Reasoning	Tool use
gpt-oss-120b	131,072	No	Yes	Yes
gpt-oss-20b	131,072	No	Yes	Yes
Qwen 3.8 27B	131,072	Yes, up to 3 images	Yes	Yes

Sources: Groq models, vision, reasoning, tool use.

gpt-oss-120b and gpt-oss-20b are OpenAI’s open-weight models, so Groq’s free plan is one way to use OpenAI models without paying. Qwen 3.8 27B is the one to pick if you want to ask about photos.

Is Groq private? What happens to your data

Groq says it doesn’t retain customer data for inference requests by default. The exception: inputs and outputs may be logged for up to 30 days to troubleshoot problems or investigate abuse, and you can opt out or turn on zero data retention under Data Controls in your Groq account (Groq).

Even so, a request to Groq means your message travels to Groq’s servers. For anything you’d rather keep on your device, use a model that runs on it.

How to get a Groq API key

1. Sign up at console.groq.com.
2. Open API Keys (console.groq.com/keys) and create a key.
3. Copy it and keep it private.

How to use the Groq API on iPhone, iPad and Mac

Modelbase is a free app for iPhone, iPad and Mac that runs AI models on the device and also connects to any server with an OpenAI-compatible API, which is what Groq offers (Groq).

1. Open Manage models. Tap the model’s name at the top of a chat, then Manage models.
2. Add a custom server. Tap +, then Custom Server.
3. Fill in the form:
   • Name: Groq
   • Address: https://api.groq.com/openai/v1 (type the whole address, including /openai/v1)
   • API Key: your Groq key
4. Tap Test Connection. When it says Connected, Groq’s models appear under Models on This Server. Tap openai/gpt-oss-120b, openai/gpt-oss-20b or qwen/qwen3.8-27b to add it. (You can also type an ID under Add Model by ID.)
5. Set up each model. Tap an added model to open its settings:
   • Accepts images: on for qwen/qwen3.8-27b, off for the gpt-oss models;
   • Can use tools: on, since Groq supports tool use on all its models; the model can then search the web and open pages by itself;
   • Context length: up to 131,072, the models’ context on Groq. A shorter one keeps you further from the 8,000-tokens-a-minute limit.
6. Tap Save. The models appear in the model picker under Your servers.

Your key is stored in the device’s Keychain; with iCloud sync on, your servers and keys sync to your other devices through your own iCloud.

The gpt-oss and Qwen models reason before they answer on Groq’s side. With a custom server, Modelbase shows the answer; the Instant, Medium and High thinking control is for on-device and OpenRouter models.

If something doesn’t work

• “Servers outside your network need https”: the address must start with https://.
• The key is refused: create a new key in the Groq console and paste it again.
• Requests fail after a while: you’ve probably hit a free limit. Wait for it to reset, or switch to a model on your device in the meantime.

Groq vs OpenRouter’s free models

Both are free ways to use large open models from an iPhone; they suit different needs.

	Groq free plan	OpenRouter free models
Models	3 chat models (gpt-oss-120b, gpt-oss-20b, Qwen 3.8 27B)	About 14 free text models from many makers
Daily requests	1,000 per model	50, or 1,000 after buying 10 credits
Per minute	30 requests, 8,000 tokens	20 requests
Data	Not retained by default (up to 30 days for troubleshooting)	Depends on each model’s provider
In Modelbase	Custom server	Built in, with a Free filter

If you want the most requests a day for free, Groq is the generous one; if you want to try many different models, OpenRouter’s free list is wider. See our OpenRouter free models guide.

When you outgrow the free tier

Groq’s Developer plan is pay-as-you-go with much higher limits; for gpt-oss-120b, for example, 1,000 requests a minute and 500,000 a day (Groq). Groq lists gpt-oss-120b at $0.15 per million input tokens and $0.60 per million output tokens (Groq models), as of November 2026. Upgrading needs a payment method and has no immediate charge (Groq).

Or don’t pay at all: for everyday chat, a model on your own device has no limits. Modelbase comes with LFM 2.5 (1.2B) and has 53 open models to download, and they work offline. See how to run an LLM locally on iPhone, iPad and Mac, and our overview of free LLM APIs for OpenRouter and Google’s Gemini API.

FAQ

Is the Groq API free?

Yes, Groq has a free plan. As of November 2026 it covers gpt-oss-120b, gpt-oss-20b and Qwen 3.8 27B, each with 30 requests a minute and 1,000 a day, plus speech-to-text with Whisper.

What are Groq’s rate limits on the free tier?

For each of the three chat models: 30 requests a minute, 1,000 requests a day, 8,000 tokens a minute and 200,000 tokens a day. Limits apply to your whole organization.

Do I need a credit card for Groq?

Not for the free plan. Groq asks for a payment method only to upgrade to its paid Developer plan.

Is Llama still free on Groq?

No. Llama 3.1 8B Instant and Llama 3.3 70B were removed for free and Developer users in August 2026.

Is Groq the same as Grok?

No. Groq is an inference company with its own chips and API. Grok is xAI’s chatbot.

Can I use Groq on my iPhone?

Yes. Add it in Modelbase as a custom server with the address https://api.groq.com/openai/v1 and your Groq API key.