> ## Documentation Index
> Fetch the complete documentation index at: https://reedai-07fa30f1.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Switching AI Models

> Learn when and how to switch between AI models to get the best results for different tasks

## Why Switch Models?

Different AI models excel at different tasks. Switching models lets you leverage each model's unique strengths without starting a new conversation.

<Note>
  You can switch models mid-conversation without losing context - ZeroTwo maintains the conversation history across all models.
</Note>

## How to Switch Models

<Steps>
  <Step title="Open Model Selector">
    Click the current model name in the top bar or press `Cmd/Ctrl + M`

    <Info>
      The model selector shows a badge with the current model's icon and name.
    </Info>
  </Step>

  <Step title="Browse Available Models">
    Models are organized by provider:

    * OpenAI (GPT-4o, GPT-5, O3, O4-mini)
    * Anthropic (Claude Sonnet 4, Claude 3.5)
    * Google (Gemini 2.5 Pro, Gemini Flash)
    * And 10+ more providers
  </Step>

  <Step title="Select New Model">
    Click on any model to switch immediately. The conversation context transfers automatically.
  </Step>

  <Step title="Continue Chatting">
    Your next message uses the newly selected model. Previous messages remain unchanged.
  </Step>
</Steps>

<Tip>
  Each message in your conversation shows which model generated it with a small icon badge.
</Tip>

## Model Selection Guide

### For Code Generation & Debugging

<CardGroup cols={2}>
  <Card title="GPT-4o" icon="code">
    **Best for**: General coding tasks

    * Clean, well-commented code
    * Multiple languages
    * Fast responses
    * Good debugging help
  </Card>

  <Card title="Claude Sonnet 4" icon="code-branch">
    **Best for**: Complex algorithms

    * Advanced code reasoning
    * Refactoring suggestions
    * Architecture design
    * Code reviews
  </Card>

  <Card title="O3 / O4-mini" icon="brain">
    **Best for**: Algorithm design

    * Complex logic problems
    * Mathematical algorithms
    * Performance optimization
    * Edge case handling
  </Card>

  <Card title="DeepSeek Reasoner" icon="lightbulb">
    **Best for**: Problem solving

    * Step-by-step reasoning
    * Algorithm explanation
    * Logic verification
    * Testing strategies
  </Card>
</CardGroup>

**Example workflow**:

```
1. Use GPT-4o to write initial code
2. Switch to Claude Sonnet 4 for refactoring
3. Switch to O3 for optimization
```

### For Writing & Content

<CardGroup cols={2}>
  <Card title="GPT-4o" icon="pen">
    **Best for**: General writing

    * Blog posts
    * Marketing copy
    * Emails
    * Creative content
  </Card>

  <Card title="Claude Sonnet 4" icon="file-lines">
    **Best for**: Long-form content

    * Technical documentation
    * Reports
    * Research papers
    * Detailed articles
  </Card>

  <Card title="GPT-5" icon="sparkles">
    **Best for**: Creative writing

    * Stories
    * Scripts
    * Poetry
    * Brand messaging
  </Card>

  <Card title="Mistral Large" icon="language">
    **Best for**: Multilingual

    * Translation
    * International content
    * Cultural adaptation
    * Multiple languages
  </Card>
</CardGroup>

**Example workflow**:

```
1. Use GPT-4o for initial draft
2. Switch to Claude Sonnet 4 for expansion
3. Switch to GPT-5 for creative polish
```

### For Research & Analysis

<CardGroup cols={2}>
  <Card title="Perplexity Sonar Pro" icon="magnifying-glass">
    **Best for**: Current information

    * Real-time data
    * Recent events
    * Market research
    * Cited sources
  </Card>

  <Card title="Claude Sonnet 4" icon="book">
    **Best for**: Document analysis

    * Long documents (200K+ tokens)
    * Research papers
    * Legal documents
    * Detailed summaries
  </Card>

  <Card title="Gemini 2.5 Pro" icon="chart-bar">
    **Best for**: Data analysis

    * Large datasets
    * Visual analysis
    * Trend identification
    * Insights generation
  </Card>

  <Card title="Qwen Max" icon="globe">
    **Best for**: Multilingual research

    * International sources
    * Cross-language analysis
    * Global perspectives
    * Translation
  </Card>
</CardGroup>

**Example workflow**:

```
1. Use Perplexity for latest information
2. Switch to Claude for deep analysis
3. Switch to Gemini for visualization
```

### For Multimodal Tasks

<CardGroup cols={2}>
  <Card title="Gemini 2.5 Pro" icon="image">
    **Best for**: Image understanding

    * Photo analysis
    * Chart reading
    * Visual Q\&A
    * Design feedback
  </Card>

  <Card title="GPT-4o" icon="images">
    **Best for**: General vision

    * Image descriptions
    * Document OCR
    * Diagram analysis
    * Visual reasoning
  </Card>

  <Card title="Claude Sonnet 4" icon="file-image">
    **Best for**: Document images

    * Screenshot analysis
    * Scanned documents
    * Complex diagrams
    * Multi-page PDFs
  </Card>

  <Card title="Qwen Max" icon="mountain-sun">
    **Best for**: Asian content

    * Chinese text in images
    * Asian language OCR
    * Cultural context
    * Regional images
  </Card>
</CardGroup>

### For Speed & Efficiency

<CardGroup cols={2}>
  <Card title="GPT-4o-mini" icon="bolt">
    **Speed**: Ultra-fast

    Quick responses for simple tasks
  </Card>

  <Card title="Gemini 2.5 Flash" icon="zap">
    **Speed**: Very fast

    Fast multimodal processing
  </Card>

  <Card title="Groq Llama 3.3" icon="gauge-high">
    **Speed**: Fastest

    Near-instant responses
  </Card>

  <Card title="Mistral Small" icon="forward">
    **Speed**: Fast

    Quick, efficient results
  </Card>
</CardGroup>

**When to prioritize speed**:

* Quick questions
* Simple tasks
* Brainstorming
* Iteration loops

### For Reasoning & Logic

<CardGroup cols={2}>
  <Card title="O3" icon="brain">
    **Reasoning**: Advanced

    * Complex math
    * Logic problems
    * Strategic thinking
    * Multi-step reasoning
  </Card>

  <Card title="DeepSeek Reasoner" icon="thought-bubble">
    **Reasoning**: Transparent

    * Shows thinking process
    * Step-by-step logic
    * Verification
    * Problem decomposition
  </Card>

  <Card title="Qwen Thinking" icon="head-side-brain">
    **Reasoning**: Reflective

    * Self-correction
    * Alternative approaches
    * Uncertainty handling
    * Meta-reasoning
  </Card>

  <Card title="Claude Sonnet 4" icon="lightbulb">
    **Reasoning**: Balanced

    * Clear explanations
    * Logical structure
    * Ethical reasoning
    * Nuanced analysis
  </Card>
</CardGroup>

## Model Comparison

### Context Window Sizes

| Model           | Context Window | Best For            |
| --------------- | -------------- | ------------------- |
| Claude Sonnet 4 | 200K tokens    | Very long documents |
| Gemini 2.5 Pro  | 1M+ tokens     | Massive documents   |
| GPT-4o          | 128K tokens    | Long conversations  |
| DeepSeek        | 64K tokens     | Extended reasoning  |
| Groq Llama 3.3  | 32K tokens     | Quick tasks         |

<Info>
  1 token ≈ 0.75 words. A 200K token context can handle \~150K words or \~300 pages of text.
</Info>

### Speed Comparison

**Fastest** → **Slowest**:

1. 🚀 Groq Llama 3.3 - Ultra-fast inference
2. ⚡ Gemini 2.5 Flash - Very fast multimodal
3. 🏃 GPT-4o-mini - Quick responses
4. 🚶 GPT-4o - Balanced speed
5. 🐢 O3 - Slower but thorough reasoning

### Cost Efficiency

**Most Efficient** → **Premium**:

1. 💰 GPT-4o-mini / Mistral Small
2. 💰💰 GPT-4o / Gemini Flash
3. 💰💰💰 Claude Sonnet / Gemini Pro
4. 💰💰💰💰 GPT-5 / O3 / Claude Opus

<Tip>
  Use efficient models for iteration, then switch to premium models for final output.
</Tip>

## Strategic Model Switching

### Workflow Patterns

**Pattern 1: Draft → Refine → Polish**

<Steps>
  <Step title="Draft (Fast Model)">
    ```
    Model: GPT-4o-mini
    Task: Generate initial outline or draft quickly
    ```
  </Step>

  <Step title="Refine (Balanced Model)">
    ```
    Model: GPT-4o or Claude Sonnet 4
    Task: Expand and improve content
    ```
  </Step>

  <Step title="Polish (Premium Model)">
    ```
    Model: GPT-5 or Claude Opus
    Task: Final refinement and quality check
    ```
  </Step>
</Steps>

**Pattern 2: Research → Analyze → Summarize**

<Steps>
  <Step title="Research (Web-Enabled)">
    ```
    Model: Perplexity Sonar Pro
    Task: Gather current information with citations
    ```
  </Step>

  <Step title="Analyze (Long Context)">
    ```
    Model: Claude Sonnet 4 or Gemini 2.5 Pro
    Task: Deep analysis of gathered information
    ```
  </Step>

  <Step title="Summarize (Clear Communication)">
    ```
    Model: GPT-4o
    Task: Create clear, actionable summary
    ```
  </Step>
</Steps>

**Pattern 3: Code → Review → Optimize**

<Steps>
  <Step title="Code (Fast Generation)">
    ```
    Model: GPT-4o
    Task: Write functional code quickly
    ```
  </Step>

  <Step title="Review (Quality Check)">
    ```
    Model: Claude Sonnet 4
    Task: Review for best practices and improvements
    ```
  </Step>

  <Step title="Optimize (Advanced Reasoning)">
    ```
    Model: O3 or DeepSeek Reasoner
    Task: Optimize algorithms and performance
    ```
  </Step>
</Steps>

## Model-Specific Tips

<AccordionGroup>
  <Accordion title="GPT-4o & GPT-5 Series">
    **Strengths**:

    * Excellent instruction following
    * Strong at creative tasks
    * Good code generation
    * Fast responses

    **Best prompting**:

    * Be direct and clear
    * Provide examples
    * Specify format preferences

    **Avoid**:

    * Extremely long documents (use Claude)
    * Very recent information (use Perplexity)
  </Accordion>

  <Accordion title="Claude Sonnet 4">
    **Strengths**:

    * Longest context window
    * Excellent for analysis
    * Thoughtful, nuanced responses
    * Strong ethical reasoning

    **Best prompting**:

    * Give comprehensive context
    * Ask for detailed analysis
    * Request step-by-step thinking

    **Avoid**:

    * Ultra-fast iteration needs (use GPT-4o-mini)
    * Real-time information (use Perplexity)
  </Accordion>

  <Accordion title="Gemini 2.5 Pro/Flash">
    **Strengths**:

    * Multimodal excellence
    * Massive context window
    * Fast processing
    * Good with data

    **Best prompting**:

    * Combine text with images
    * Reference multiple sources
    * Ask for visual analysis

    **Avoid**:

    * Very specialized coding tasks
    * Pure text when speed is critical
  </Accordion>

  <Accordion title="O3 & Reasoning Models">
    **Strengths**:

    * Complex problem solving
    * Mathematical reasoning
    * Logic and proofs
    * Strategic thinking

    **Best prompting**:

    * Present complex problems
    * Ask for step-by-step solutions
    * Request multiple approaches

    **Avoid**:

    * Simple, quick tasks
    * Creative writing
    * General questions
  </Accordion>

  <Accordion title="Perplexity Sonar Pro">
    **Strengths**:

    * Real-time web access
    * Cited sources
    * Current events
    * Fact-checking

    **Best prompting**:

    * Ask for recent information
    * Request citations
    * Compare current data

    **Avoid**:

    * Historical analysis
    * Creative tasks
    * Code generation
  </Accordion>
</AccordionGroup>

## Setting a Default Model

Choose your preferred starting model:

1. Go to **Settings** > **Preferences**
2. Find **Default Model** section
3. Select your preferred model
4. Save changes

<Info>
  Your default model is used for new conversations. You can still switch to any model at any time.
</Info>

## Model Availability by Subscription

### Free Tier

* GPT-4o-mini
* Gemini 2.5 Flash
* Mistral Small
* Limited messages per day

### Pro Tier (\$20/month)

* All Free tier models
* GPT-4o, GPT-4.1
* Claude 3.5 Sonnet/Haiku
* Gemini 2.5 Pro
* Most models unlimited

### Premium Tier (\$40/month)

* All Pro tier models
* GPT-5, O3
* Claude Sonnet 4, Opus 4
* Unlimited usage
* Priority access

<Tip>
  Start with Free tier to learn model differences, then upgrade based on your most-used models.
</Tip>

## Troubleshooting

<AccordionGroup>
  <Accordion title="Model not responding">
    **Check**:

    * Internet connection
    * Platform status
    * Subscription tier for premium models
    * Try switching to another model
  </Accordion>

  <Accordion title="Context not transferring">
    **Solution**:

    * Context transfers automatically
    * Try refreshing the page
    * Check if conversation is too long
    * Start a new chat if needed
  </Accordion>

  <Accordion title="Different results from same prompt">
    **This is normal**:

    * Different models have different capabilities
    * Each model has unique strengths
    * Results vary by design
    * Use this to your advantage!
  </Accordion>
</AccordionGroup>

## Next Steps

<CardGroup cols={2}>
  <Card title="Model Providers" icon="robot" href="/overview/models-and-providers">
    Learn about all available AI providers
  </Card>

  <Card title="Upload Files" icon="file" href="/getting-started/files-and-history">
    Work with documents in your conversations
  </Card>

  <Card title="Advanced Features" icon="wand-sparkles" href="/tools/canvas">
    Explore Canvas, Web Search, and more
  </Card>

  <Card title="Best Practices" icon="lightbulb" href="/prompts/overview">
    Master effective prompting techniques
  </Card>
</CardGroup>

<Check>
  Experiment with different models to discover which works best for your specific needs!
</Check>