AI Model Selection for Teacher Evaluations

Overview

PrepareBuddy provides multiple AI models for evaluating teacher feedback. You can select different models based on your evaluation needs and preferences.

Available AI Models

The following AI models are available for teacher evaluations:

1. Qwen 3 32B (Default)

• Model ID: qwen/qwen3-32b
• Best for: Comprehensive teacher evaluation with detailed analysis
• Tokens: 4,000 maximum
• Temperature: 0.3 (balanced creativity)

2. Llama 3.3 70B Versatile

• Model ID: llama-3.3-70b-versatile
• Best for: Versatile evaluations with strong reasoning capabilities
• Tokens: 4,000 maximum
• Temperature: 0.3 (balanced creativity)

3. Llama 4 Maverick (Vision)

• Model ID: meta-llama/llama-4-maverick-17b-128e-instruct
• Best for: Evaluating submissions with visual content (images, diagrams)
• Tokens: 4,000 maximum
• Temperature: 0.3 (balanced creativity)
• Special: Supports image analysis

4. DeepSeek R1 Distill 70B

• Model ID: deepseek-r1-distill-llama-70b
• Best for: Deep analysis and identifying subtle teaching patterns
• Tokens: 4,000 maximum
• Temperature: 0.3 (balanced creativity)

5. Llama 4 Scout

• Model ID: meta-llama/llama-4-scout-17b-16e-instruct
• Best for: Quick evaluations with good balance of speed and quality
• Tokens: 4,000 maximum
• Temperature: 0.3 (balanced creativity)

6. Mistral Saba 24B

• Model ID: mistral-saba-24b
• Best for: Balanced performance across various evaluation criteria
• Tokens: 4,000 maximum
• Temperature: 0.3 (balanced creativity)

How to Select AI Models

When Evaluating Teacher Feedback

1. Access Teacher Evaluation: Go to Assessment → Teacher Evaluation → Evaluate Feedback
2. Select AI Model: Look for the "Choose AI Model" dropdown
3. View Model Details: Hover over models to see descriptions and capabilities
4. Make Selection: Choose the model that best fits your evaluation needs
5. Complete Evaluation: The selected model will be used for the evaluation

Model Selection Tips

• For text-only feedback: Use Qwen 3 32B (default) or Llama 3.3 70B
• For feedback with images: Use Llama 4 Maverick (Vision)
• For quick evaluations: Use Llama 4 Scout
• For detailed analysis: Use DeepSeek R1 Distill 70B
• For balanced performance: Use Mistral Saba 24B

Model Features

Global Models

All models are configured as global models, meaning they are available to all users across the platform.

Model Settings

• Max Tokens: 4,000 tokens for comprehensive responses
• Temperature: 0.3 for balanced creativity and consistency
• Vision Support: Only Llama 4 Maverick supports image analysis

Session Persistence

Your selected model is remembered during your session, so you don't need to reselect it for each evaluation.

Troubleshooting

Model Not Available

If a model is not available: - The model may be temporarily disabled - Your organization may have restricted access - Contact your administrator for assistance

Evaluation Issues

If you encounter problems: - Try a different model - Check your internet connection - Contact support if issues persist

Getting Help

For questions about AI model selection: - Email: support@preparebuddy.com - Documentation: Check the Assessment System documentation - Administrator: Contact your organization administrator

The AI model selection feature helps you choose the best evaluation approach for different types of teacher feedback and assessment scenarios.