Skip to main content

Research Mode

(Experimental Feature — enable in Profile > Settings > Experimental Features)

Research Mode lets you compare responses from multiple AI models side-by-side. Ask a question once and see how different models answer it, helping you choose the best model for your needs.

Overview​

Why Use Research Mode?​

  • Compare models: See how Claude, GPT-4, Gemini, and others respond differently
  • Test parameters: Try the same model with different temperature settings
  • Make informed choices: Find which model works best for specific tasks
  • Save time: No need to ask the same question multiple times

How It Works​

  1. Enable Research Mode in your notebook settings
  2. Configure up to 4 different model configurations
  3. Send your prompt
  4. All models respond simultaneously in a grid view
  5. Compare results side-by-side

Setting Up Research Mode​

Enabling the Feature​

  1. Open AI Settings in your notebook
  2. Find the Tools section
  3. Toggle Research Mode on
  4. Configure your model selections

Configuring Models​

For each of the 4 slots, you can set:

SettingDescription
ModelChoose any available AI model
TemperatureControl response creativity (0.0 - 1.0+)
Max TokensLimit response length

Example Configurations​

Compare Model Families:

  • Slot 1: Claude 3.5 Sonnet
  • Slot 2: GPT-4o
  • Slot 3: Gemini 2.5 Pro
  • Slot 4: Llama 3

Test Temperature Settings:

  • Slot 1: Claude (temperature 0.0 - focused)
  • Slot 2: Claude (temperature 0.5 - balanced)
  • Slot 3: Claude (temperature 0.8 - creative)
  • Slot 4: Claude (temperature 1.0 - very creative)

Using Research Mode​

Sending Prompts​

When Research Mode is active:

  1. Type your prompt normally
  2. Press Enter to send
  3. All configured models receive the same prompt
  4. Responses stream in simultaneously

Viewing Results​

Results appear in a grid layout:

  • Desktop: 2x2 grid showing all 4 responses
  • Mobile: Stacked cards you can scroll through
  • Each response shows the model name and settings

Comparing Responses​

Look for differences in:

  • Tone and style: How each model writes
  • Detail level: Some models are more thorough
  • Accuracy: Fact-checking across models
  • Creativity: Unique approaches to problems
  • Speed: Which models respond faster

Best Practices​

When to Use Research Mode​

Good use cases:

  • Evaluating models for a new project
  • Finding the best model for creative writing vs. technical tasks
  • Testing how models handle specific domains
  • Comparing cost vs. quality trade-offs

Not ideal for:

  • Quick single questions (use normal mode)
  • Very long conversations (uses 4x credits)
  • Simple factual lookups

Optimizing Your Comparisons​

  1. Use consistent prompts: Ask the exact same question to all models
  2. Test multiple times: Run a few comparisons to see patterns
  3. Note the differences: Keep track of which model excels at what
  4. Consider cost: Research Mode uses credits for all 4 models

Understanding Results​

What to Look For​

AspectWhat It Tells You
Response timeModel speed and efficiency
LengthVerbosity preferences
FormatHow each structures answers
AccuracyFactual correctness
HelpfulnessPractical usefulness

Making Decisions​

After comparing:

  • Note which model performed best for your use case
  • Consider switching your default model
  • Save preferred configurations for future use

Cost Considerations​

Research Mode uses significantly more credits:

  • 4x credit usage per prompt (one for each model)
  • Higher costs for premium models
  • Consider using smaller prompts for initial testing

Tip: Use Research Mode sparingly for evaluation, then switch to your preferred single model for regular use.

Limitations​

Current Restrictions​

  • Maximum 4 model configurations
  • All models receive identical context
  • Some models may not be available depending on your plan
  • Long conversations can become expensive

Coming Soon​

  • Save and load configuration presets
  • Export comparison results
  • Automated quality scoring
  • Team-shared configurations