# Instant Inference with Groq

Groq's LPU (Language Processing Unit) delivers inference speeds that feel like magic. Here's how to get 500+ tok/s with Ishi.

## Why Groq?

- **Blazing fast** — 500+ tokens/second
- **Free tier** — Generous rate limits
- **Open models** — Llama, Mixtral, Gemma
- **Real-time feel** — Responses appear instantly

## Quick Setup

```json
{
  "provider": "groq",
  "model": "llama-3.3-70b-versatile"
}
```

Get your API key at [console.groq.com](https://console.groq.com/).

## Speed Comparison

\| Provider \| Speed \| Model \|
\|----------\|-------\|-------\|
\| Groq \| 500+ tok/s \| Llama 3.3 70B \|
\| OpenAI \| 50-80 tok/s \| GPT-4o \|
\| Anthropic \| 40-60 tok/s \| Claude Sonnet \|
\| Ollama (local) \| 15-40 tok/s \| Llama 3.2 \|

## Perfect For

- **Iterative workflows** — Rapid back-and-forth
- **Real-time processing** — Live document streaming
- **High-volume tasks** — Batch file processing
- **Development** — Quick testing and iteration

---

## Try Ishi Today

Download Ishi and start automating your workflow with the Glass Box philosophy.

[Download Free](/content/download/index.html)