> ## Documentation Index
> Fetch the complete documentation index at: https://docs.stenox.app/llms.txt
> Use this file to discover all available pages before exploring further.

# Groq LLMs

> Ultra-fast AI enhancement with Llama, Qwen, and DeepSeek on Groq infrastructure

"og:title": "Stenox Docs"

<Warning>
  This is an **AI ENHANCEMENT** provider. It improves transcribed text with grammar correction and formatting.
  For speech-to-text transcription, see [Transcription Providers](/providers/overview#transcription-providers).
</Warning>

## Overview

Groq provides blazing-fast inference for open-source large language models using their LPU (Language Processing Unit) architecture. This enables some of the fastest AI enhancement available with excellent quality.

**Key Benefits:**

* ⚡ **Blazing fast** - Sub-second enhancement with Groq's LPU
* 🎯 **High quality** - Llama 3.3/4, Qwen 3, DeepSeek R1 models
* 🔄 **Same API key** - Use for both Groq Whisper and Groq LLMs
* 🎁 **Free tier** - Generous free usage with rate limits

<Tip>
  **Perfect combo:** Use [Groq Whisper](/providers/transcription/groq-whisper) for transcription + Groq LLMs for enhancement with a **single API key**!
</Tip>

## Free Tier

<Tip>
  **Groq offers a generous free tier** with rate limits instead of credit limits.

  **What you get:**

  * **Free tier with rate limits** (varies by model)
  * **All models included** - Llama, Qwen, DeepSeek R1
  * **Same API for Whisper** - One key for transcription + enhancement
  * **No expiration** - Free tier doesn't expire

  Perfect for regular users who want fast, quality enhancement without costs.
</Tip>

### Free Tier Limits

Groq's free tier varies by model. Generally:

* **Rate limits** instead of credit limits
* **Tokens per minute** limits per model
* **Sufficient for personal use** and many professional use cases
* **Check [console.groq.com](https://console.groq.com)** for current limits

<Note>
  When you hit rate limits, either wait a few minutes or use a fallback profile with [MLX (local)](/providers/llm/mlx-local).
</Note>

## Getting Your Free API Key

<Steps>
  <Step title="Visit Groq Console">
    Go to [console.groq.com](https://console.groq.com) in your browser.
  </Step>

  <Step title="Sign up for free">
    Click **Sign Up** and create an account with:

    * Email and password
    * Or sign in with Google/GitHub

    **No credit card required!**
  </Step>

  <Step title="Verify your email">
    Check your email for verification link from Groq and click it.
  </Step>

  <Step title="Create API key">
    Once logged in:

    1. Go to **API Keys** section
    2. Click **Create API Key**
    3. Give it a name (e.g., "Stenox")
    4. Copy the key immediately (can't view again!)
  </Step>

  <Step title="Add key to Stenox">
    1. Open Stenox Settings
    2. Go to **AI Enhancement** tab
    3. Select **Groq** as provider
    4. Paste your API key
    5. Choose your model (Llama 3.3 70B or DeepSeek R1 recommended)
    6. Click **Save**
  </Step>
</Steps>

<Note>
  **One API key for everything:** This same key works for [Groq Whisper (transcription)](/providers/transcription/groq-whisper) and Groq LLMs (enhancement). Configure once!
</Note>

## Available Models

Groq offers multiple open-source LLMs optimized for their LPU hardware:

### Llama 3.3 70B Versatile (Recommended)

**Best all-around model:**

* Excellent grammar correction and formatting
* Balanced speed and quality
* Versatile for all use cases
* Good context understanding

**Best for:** Default choice, everyday enhancement

### DeepSeek R1 Distill Llama 70B

**Reasoning-focused model:**

* Excellent for complex edits
* Strong logical reasoning
* Very good at restructuring
* Slightly slower than Llama 3.3

**Best for:** Complex text improvements, professional writing

### Llama 4 Scout 17B

**Faster, smaller model:**

* Faster inference than 70B models
* Good quality for size
* Lower resource usage
* Good for quick enhancement

**Best for:** Speed-critical use cases

### Qwen 3 72B

**Multilingual specialist:**

* Excellent for non-English languages
* Good grammar correction
* Strong formatting abilities

**Best for:** Multilingual enhancement

<Tip>
  **Recommendation:** Start with **Llama 3.3 70B Versatile** for the best balance. Try **DeepSeek R1** for maximum quality.
</Tip>

## Configuration in Stenox

<Steps>
  <Step title="Open Stenox Settings">
    Click Stenox menu bar icon → **Settings**
  </Step>

  <Step title="Go to AI Enhancement tab">
    Select the **AI Enhancement** tab
  </Step>

  <Step title="Select Groq">
    Choose **Groq** from the provider dropdown
  </Step>

  <Step title="Enter API key">
    Paste your API key from console.groq.com

    (Same key as Groq Whisper if you use it!)
  </Step>

  <Step title="Choose model">
    Select **Llama 3.3 70B Versatile** (recommended)
  </Step>

  <Step title="Optional: Custom prompt">
    Configure enhancement behavior (see Custom Prompts section)
  </Step>

  <Step title="Save">
    Click **Save** to activate Groq LLMs
  </Step>
</Steps>

## What Groq LLMs Do

### Grammar Correction

* Comprehensive grammar fixes
* Subject-verb agreement
* Tense consistency
* Article and preposition corrections
* Complex sentence structure improvements

**Example:**

* **Before:** "me and my team was working on the project all day yesterday and we finally finished it"
* **After:** "My team and I were working on the project all day yesterday, and we finally finished it."

### Formatting & Punctuation

* Professional punctuation
* Proper capitalization
* Number and date formatting
* List structuring
* Paragraph organization

**Example:**

* **Before:** "our meeting is scheduled for march 15th at 2pm please bring the q4 reports"
* **After:** "Our meeting is scheduled for March 15th at 2 PM. Please bring the Q4 reports."

### Tone Adjustment

With custom prompts, Groq LLMs excel at tone modification:

* **Professional tone** - Formal, business-appropriate
* **Casual tone** - Friendly, conversational
* **Technical tone** - Precise, documentation-style
* **Persuasive tone** - Engaging, convincing

### Advanced Capabilities

**DeepSeek R1** especially excels at:

* Restructuring for clarity
* Making implicit information explicit
* Improving logical flow
* Complex reasoning-based edits

## Performance

**Expected enhancement times with Groq LLMs:**

| Transcription Length       | Enhancement Time |
| -------------------------- | ---------------- |
| Short (1-2 sentences)      | \< 0.5 seconds   |
| Medium (paragraph)         | \< 1 second      |
| Long (multiple paragraphs) | 1-2 seconds      |

<Note>
  Groq's LPU architecture makes enhancement nearly instantaneous, even for large models like 70B parameter LLMs.
</Note>

## Custom Prompts

Configure how Groq LLMs enhance your text in Profile settings:

<Tabs>
  <Tab title="Default">
    ```
    Fix grammar and punctuation errors.
    Keep the original meaning and tone.
    ```

    **Use for:** General dictation
  </Tab>

  <Tab title="Professional Email">
    ```
    Fix grammar and rewrite in a professional, business-appropriate tone.
    Make it clear, concise, and action-oriented.
    Remove filler words and redundancy.
    ```

    **Use for:** Work emails, business communications
  </Tab>

  <Tab title="Technical Documentation">
    ```
    Fix grammar and format for technical documentation.
    Preserve all technical terms exactly as stated.
    Use active voice, clear language, and proper formatting.
    Add appropriate section breaks if needed.
    ```

    **Use for:** Code documentation, technical writing
  </Tab>

  <Tab title="Creative Writing">
    ```
    Fix grammar while preserving creative voice.
    Enhance descriptive language where appropriate.
    Maintain the author's style and tone.
    ```

    **Use for:** Blog posts, articles, creative content
  </Tab>
</Tabs>

## Usage Tracking

Monitor your Groq usage:

1. Visit [console.groq.com](https://console.groq.com)
2. Go to **Usage** section
3. View:
   * API calls made
   * Tokens consumed
   * Rate limit status
   * Current tier

<Note>
  Stenox does not track Groq usage. Check console.groq.com for accurate statistics.
</Note>

## Privacy Considerations

**What Groq sees:**

* Your transcribed text is sent to Groq servers
* Text may be logged for service improvement
* Data is encrypted in transit (HTTPS)
* Review Groq's privacy policy for details

**For complete privacy:**

* Use [MLX (local)](/providers/llm/mlx-local) for 100% offline enhancement

## When to Use Groq LLMs

<CardGroup cols={2}>
  <Card title="Ultra-fast processing" icon="bolt">
    LPU-accelerated inference for near-instant results.
  </Card>

  <Card title="High quality" icon="medal">
    70B parameter models with excellent capabilities.
  </Card>

  <Card title="Groq ecosystem" icon="link">
    Same API key as Groq Whisper - seamless integration.
  </Card>

  <Card title="Free tier" icon="gift">
    Generous free usage for personal and professional use.
  </Card>
</CardGroup>

## Groq Ecosystem: Full Pipeline

Use Groq for **both** transcription and enhancement:

### Complete Groq Setup

<Steps>
  <Step title="Sign up once">
    Create one Groq account at console.groq.com
  </Step>

  <Step title="One API key">
    Generate a single API key for all Groq services
  </Step>

  <Step title="Configure transcription">
    Use Groq Whisper (whisper-large-v3-turbo) for transcription
  </Step>

  <Step title="Configure enhancement">
    Use Groq LLMs (Llama 3.3 70B or DeepSeek R1) for enhancement
  </Step>

  <Step title="Done!">
    One provider, one API key, complete pipeline
  </Step>
</Steps>

**Benefits:**

* Single account management
* Unified billing/usage tracking
* Ultra-fast end-to-end processing (\~2 seconds total)
* Simple configuration

[Learn about Groq Whisper →](/providers/transcription/groq-whisper)

## Troubleshooting

<AccordionGroup>
  <Accordion title="API key not working" icon="key">
    * Ensure you copied the complete key (no spaces)
    * Verify account is active (check email verification)
    * Try creating a new API key
    * Check for typos
  </Accordion>

  <Accordion title="Rate limit exceeded" icon="gauge-high">
    * You've hit free tier rate limits
    * Wait a few minutes and try again
    * Use [MLX (local)](/providers/llm/mlx-local) as backup
    * Create a profile without enhancement for quick notes
    * Upgrade to paid tier for higher limits
  </Accordion>

  <Accordion title="Enhancement fails" icon="triangle-exclamation">
    * Check internet connection
    * Verify API key is valid in Groq console
    * Check Groq status page for service issues
    * Try a shorter test enhancement
  </Accordion>

  <Accordion title="Quality not as expected" icon="spell-check">
    * Try DeepSeek R1 for better reasoning
    * Adjust custom prompt in Profile settings
    * Provide more specific enhancement instructions
    * Ensure transcription quality is good (can't fix bad transcription)
  </Accordion>
</AccordionGroup>

## Pricing After Free Tier

Groq offers competitive pay-as-you-go pricing when you exceed free tier:

* **Pricing varies by model** - Check [console.groq.com](https://console.groq.com)
* **Very affordable** - Among the cheapest cloud LLM options
* **No commitment** - Pay only for usage

<Note>
  Most users stay within free tier limits. Check Groq console for current pricing.
</Note>

## Model Comparison

| Model                 | Speed     | Quality   | Best For                 |
| --------------------- | --------- | --------- | ------------------------ |
| **Llama 3.3 70B**     | Very Fast | Excellent | General use, balanced    |
| **DeepSeek R1 70B**   | Fast      | Superior  | Complex edits, reasoning |
| **Llama 4 Scout 17B** | Fastest   | Very Good | Speed-critical           |
| **Qwen 3 72B**        | Very Fast | Excellent | Multilingual             |

## Next Steps

<CardGroup cols={2}>
  <Card title="Groq Whisper" icon="microphone" href="/providers/transcription/groq-whisper">
    Complete the Groq pipeline with Groq Whisper transcription.
  </Card>

  <Card title="Create Profiles" icon="layer-group" href="/profiles/creating-profiles">
    Different settings for different apps and use cases.
  </Card>

  <Card title="Free Tier Guide" icon="gift" href="/guides/free-tier-providers">
    Maximize free usage across all providers.
  </Card>

  <Card title="Cloud Setup" icon="cloud" href="/guides/cloud-setup">
    Recommended cloud configuration with Groq.
  </Card>
</CardGroup>