How do I choose the right AI model in Jylo?

Quick Answer

For most everyday tasks, we recommend starting with o3-mini as it offers a great balance of speed, reasoning ability, and cost. If you're analysing multiple large documents at once, GPT-4.1 is your best bet with its massive context window.

What Are AI Models and Why Do They Matter?

AI models are the "brains" behind Jylo's analysis capabilities. Different models excel at different tasks—some are lightning-fast but less detailed, while others take more time but provide deeper insights.

Choosing the right model means getting the best balance of speed, accuracy, and cost for your specific needs. The good news? You can always switch between models to find what works best for your documents!

Our Available Models

Here's a breakdown of the AI models currently available in Jylo:

o3-mini

• Best for: Everyday document analysis and STEM-related tasks
• Context Window: 200K tokens (~150K words)
• Strengths:
  • Excellent balance of reasoning and efficiency
  • Great at scientific and domain specific content
• Limitations: Not as powerful as larger models for very complex tasks
• Speed: Fast to medium
• Cost: ££ (Medium cost)

GPT-4o

• Context Window: 128K tokens (~90K words)
• Strengths:
  • Excellent for multilingual content
  • Very fast responses
• Limitations: May sacrifice some depth for speed with complex reasoning
• Speed: Fast
• Cost: ££ (Medium cost)
• Tip: Great for executing simple to medium complexity Flows on large datasets

o1

• Best for: Complex problems requiring careful step-by-step reasoning
• Context Window: 128K tokens (~90K words)
• Strengths:
  • Excels at STEM fields
  • Works through problems methodically
  • Top performance on safety benchmarks
• Limitations: Takes longer to respond due to thorough reasoning process
• Speed: Medium to slow
• Cost: £££ (Higher cost)

GPT-4.1

• Best for: Handling extremely large documents or multiple documents at once
• Context Window: 1 million tokens (~750K words)
• Strengths:
  • Massive context window (up to 1 million tokens)
  • Nuanced language understanding
• Limitations: Response time increases with larger documents
• Speed: Medium (slower with very large documents)
• Cost: ££ (Medium cost)
• Tip: Perfect for processing many documents with Assistant or very large documents in a Flow

GPT-4.5-preview

• Best for: Natural, conversational interactions and precision analysis
• Context Window: 128K tokens (~90K words)
• Strengths:
  • More human-like conversations
  • Lowest hallucination rate
  • Superior instruction following
• Limitations: As a preview, may have some inconsistencies
• Speed: Medium
• Cost: £££ (Higher cost)
• Tip: Excellent choice when factual precision is paramount

Llama 3.3

• Best for: General text processing with good value for money
• Context Window: 128K tokens (~90K words)
• Strengths:
  • Exceptional instruction following
  • Good multilingual capabilities
• Speed: Medium
• Cost: £ (Lower cost)
• Tip: Great choice when you need precise following of analytical instructions

Mistral Large 2411

• Best for: Enterprise applications and multilingual document analysis
• Context Window: 128K tokens (~90K words)
• Strengths:
  • Excellent multilingual capabilities supporting dozens of languages
• Speed: Medium
• Cost: ££ (Medium cost)
• Tip: Ideal for analysing international document collections

Phi 3 Medium

• Best for: Efficient processing in resource-limited environments
• Strengths:
  • Good performance in constrained computing environments
• Limitations: Not as powerful as larger models for complex tasks
• Speed: Fast
• Cost: £ (Lower cost)
• Tip: Great option when you need quick responses for straightforward tasks

Phi 4

• Best for: Edge computing and on-device applications
• Strengths:
  • Improved multilingual support over Phi 3
• Limitations: Limited by smaller size compared to larger models
• Speed: Fast
• Cost: £ (Lower cost)
• Tip: Excellent choice when privacy or connectivity is a concern

Deepseek R1

• Best for: Advanced mathematical and coding problems
• Strengths:
  • Specialized for complex reasoning tasks
  • Excellent at step-by-step problem solving
  • Massive parameter count for sophisticated analysis
• Speed: Medium to slow
• Cost: £ (Lower cost)
• Tip: For best results with mathematical problems, ask it to reason step-by-step

Our Recommendations

We've tested all these models extensively, and here's what we suggest:

• For everyday document analysis: o3-mini gives you the best balance of speed, reasoning ability, and cost for most common tasks.
• For analysing multiple large documents: GPT-4.1 is your go-to with its massive context window that can handle up to 1 million tokens.
• For budget-conscious users: Deepseek R1 offers excellent reasoning capabilities at a lower cost point.

Common Questions

How do I change which model I'm using?

Click the model dropdown in the top-left corner of the Assistant interface, then select your preferred model.

Will switching models affect my previous analyses?

No, changing models only affects new analyses. Your previous work remains intact.

Do different models cost different amounts?

Yes, generally more powerful models cost more credits. You can see the relative cost indicators (£, ££, £££) in the model descriptions above.

If You're Still Unsure...

Not sure which model to choose? Here's our suggestion:

1. Start with o3-mini for your first few analyses
2. Try GPT-4.1 if you're working with very large documents
3. Consider Deepseek R1 if you're budget-conscious but need good reasoning capabilities

Still need help deciding? Contact us via Zendesk.