Groq builds its own inference chips (LPUs) designed only for language models, rather than repurposing GPUs. The result is deterministic, extremely high token throughput — which you experience in Pluely as answers that appear nearly instantly.

Is my Groq API key safe in Pluely?

Yes. Keys are encrypted and stored locally on your device — they never leave your computer except in requests to Groq itself. Pluely's servers never see them.

Does Pluely charge for using Groq?

No. Custom providers are unmetered and fully covered by the free plan — you pay only Groq for your own usage.

Which Groq models can I use?

Any model the endpoint serves — for example llama-4-maverick, qwen3-32b, deepseek-r1-distill-llama-70b. You choose which ones appear in the overlay's picker.

Can I use Groq alongside Pluely's bundled models?

Yes. On Pro plans the picker shows your custom providers next to the 200+ bundled models; on the free plan your own providers and CLIs power everything.

Groq, in an invisible overlay.

Groq's custom LPU hardware streams hundreds of tokens per second — connected to Pluely, full answers appear almost before you finish asking. For live meetings, where every second of silence counts, raw speed is a feature.

Download free Read the setup guide

Your key, encrypted on-deviceUnmetered — free plan covers itOpenAI-compatible

All custom providers

Why Groq?

The fastest inference on the planet.

Speed you can feel

Groq routinely streams 300-800 tokens per second — 5-10x typical GPU serving. In Listen mode the suggested answer is fully written before a normal provider would have produced its first sentence.

Free tier for real use

Groq's free tier has day-one usable limits — enough to run Pluely through a normal day of questions before you ever add a card.

Strong open models, zero setup

Llama 4, Qwen 3, DeepSeek distills — the best open models, already deployed and tuned. You get open-model economics with big-lab convenience.

Connection details

Base URL: https://api.groq.com/openai/v1
Get a key: console.groq.com
Example models: llama-4-maverickqwen3-32bdeepseek-r1-distill-llama-70b

Connect Groq in three steps

Any OpenAI-compatible endpoint works the same way.

Get your Groq API key

Create a key at console.groq.com. It stays on your device — Pluely encrypts it locally and never sends it anywhere except Groq itself.

Add it as a custom provider

Open Pluely's dashboard → AI Providers → Add Custom Provider. Paste the base URL (https://api.groq.com/openai/v1) and your key, and pick the models you want available.

Pick it and ask

Your Groq models now appear in the overlay's model picker. Select one and every Ask, screenshot, and meeting answer runs on your key — unmetered by Pluely.

Everything Pluely does, on Groq

Connecting a provider doesn't limit the product — every mode routes through it.

Ask Groq about your screen

Capture the full screen, drag-select just the error or the clause that matters, attach PDFs and documents with built-in OCR, or turn on Use image so every message carries a fresh screenshot. Groq answers with the actual context in front of you.

Groq answers your meetings live

Listen mode transcribes your mic and system audio in real time with speaker labels, and Groq drafts the suggested answer before the question finishes — automatically on every question, after each pause, or only when you tap Suggest.

Summaries, transforms, and notes

One tap turns any answer, transcript, or document into a summary, key points with action items, a translation, or a longer draft — and meeting notes build themselves as you go. All of it runs on Groq.