Groq is an inference platform that runs open AI models with very low latency. It has no connection to Grok, from xAI.
2026-08-19
Groq is a native integration of NicoChat, with its own screen under Integrations > Artificial Intelligence > Groq, alongside OpenAI, Google Gemini, Claude AI, X AI, DeepSeek and Coze. You connect your Groq API key and gain one more AI provider available in flows, agents and AI tasks.
Groq is not Grok
They are different companies with different products. Groq (with a q) is an inference infrastructure company founded in 2016 — it has no model of its own, it runs open models very fast. Grok (with a k) is the language model from xAI, Elon Musk’s company, and in NicoChat it lives in the X AI integration. A key from one will not work on the other.
What Groq is for
Groq sells speed, not intelligence. It runs models on a chip of its own, the LPU (Language Processing Unit), designed for inference only — the result is a response that arrives in a fraction of the time the same model would take on an ordinary GPU, in the range of hundreds of tokens per second.
The catalog is made of open models (open-weight), such as those from the Llama, Qwen, Kimi and GPT-OSS families. There is no GPT-5, Claude or Gemini on Groq: those are offered by each one’s own integration.
Where that trade-off pays off in NicoChat:
Voice and phone support, where every second of waiting is silence on the line.
Flows with several AI calls in sequence — detect intent, classify, extract one field and only then answer. The delay of each step adds up, and that is where latency shows.
Short, repetitive tasks: classifying a message, deciding which queue to send the contact to, summarizing a conversation, translating a sentence.
High volume on a tight budget, since open models cost far less per token than the closed top-tier models.
And where it does not pay off: long, hard reasoning, finely polished marketing copy, enormous context. For that, prefer OpenAI, Claude AI or Google Gemini.
Groq × Grok: which is which
Groq
Grok
Who it is
A chip and inference cloud company, founded in 2016 by Jonathan Ross
The language model from xAI, Elon Musk’s company, launched in 2023
What it delivers
Runs third-party open models with very low latency
It is the model itself, with reasoning, vision and very long context
Click Create API Key, give it a name and copy the generated value — it is shown only once. If you lose it, generate another one.
Groq has a free plan
You can test it without a card, within limits on requests and tokens per minute. For production, enable billing in the Groq console itself — NicoChat does not handle that payment, usage is billed directly between you and Groq.
Installing it in NicoChat
1
Open the integration
In your workspace dashboard, go to Integrations > Artificial Intelligence > Groq.
2
Check the base URL
Keep the default value https://api.groq.com/openai/v1. Only change it if you run your own Groq instance or are going to point to a different compatible endpoint.
3
Paste the key
Fill the API Key field with the value you copied from the console.
4
Save
Click Save. From then on Groq appears as a provider in the AI screens.
The Groq API is compatible with the OpenAI format — that is why the base URL ends in /openai/v1. It does not mean you are talking to OpenAI: the call goes to Groq’s servers, just in the same request format.
Where Groq shows up once connected
In the flow builder’s Action block: create an action node, open the Integrations tab and scroll down to Groq.
In AI Agents, by selecting Groq on the settings tab. The default model is llama-3.1-8b-instant.
Usage shows up normally in the AI usage logs, alongside the other providers.
Actions available with Groq
Inside the Action block, NicoChat offers these Groq actions:
Action
What it is for
Create Chat Completions
The text answer itself. It is the equivalent of OpenAI’s chat completion.
Clear remembered chat history
Erases the stored context from previous conversations with that contact. Less context = fewer tokens and a faster answer.
List models
Returns the models available in your Groq account — handy for finding the exact name of a new model.
Vision — image to text
Analyzes an image that was sent and answers in text, following the prompt you wrote.
Text to speech
Converts a block of text into audio.
Speech to text
Transcribes an incoming audio — the classic case of a WhatsApp voice note turning into text.
Translate audio into English
Transcribes the audio already translated into English, in a single call.
Frequently asked questions
No. The providers coexist. You can keep OpenAI for the answers that demand more quality and use Groq for the fast steps of the same flow.
Yes. The integration is native and costs nothing in NicoChat, but API usage is billed by Groq, on the account you created there.
No. They are different services. The xAI key goes in the X AI integration; the Groq one goes on this screen.
Yes. Besides text, the Action block offers vision (image to text), text to speech, speech to text and audio translation into English — the table above lists them all.
In AI Agents, llama-3.1-8b-instant. You can switch to whichever you prefer; use the List models action to see what is available on your account.
