Skip to content

models.md

Latest commit

 

History

History
162 lines (125 loc) · 3.35 KB

models.md

File metadata and controls

162 lines (125 loc) · 3.35 KB

Model Configuration Guide

Available Models

GPT-4 Models

  • gpt-4: Latest GPT-4 model, best for complex reasoning
  • gpt-4-turbo: Faster version with good performance
  • gpt-4-32k: Extended context window (32,768 tokens)

GPT-3.5 Models

  • gpt-3.5-turbo: Fast and efficient for most tasks
  • gpt-3.5-turbo-16k: Extended context window (16,384 tokens)

Setting Your Default Model

Command Line

# Set GPT-4 as default
gh copilot config set model gpt-4

# Set GPT-3.5 for faster responses
gh copilot config set model gpt-3.5-turbo

# Use extended context version
gh copilot config set model gpt-4-32k

Environment Variable

export GH_COPILOT_MODEL="gpt-4"

Configuration File

Edit ~/.config/gh/copilot/config.yml:

default_model: gpt-4

Model Selection Guidelines

Use GPT-4 for:

  • Complex code architecture decisions
  • Detailed code reviews
  • Advanced debugging
  • Learning new technologies
  • Writing documentation

Use GPT-3.5 for:

  • Quick code snippets
  • Simple explanations
  • Basic debugging
  • Faster iteration
  • Testing ideas

Context Windows

Understanding Token Limits

# Check current model's context window
gh copilot info context

# View token usage for conversation
gh copilot usage --conversation

Model Context Sizes

Model Context Window Best For
GPT-4 8,192 tokens Standard tasks
GPT-4-32K 32,768 tokens Large codebases
GPT-3.5-Turbo 4,096 tokens Quick tasks
GPT-3.5-Turbo-16K 16,384 tokens Medium projects

Managing Context

# Clear conversation to free tokens
gh copilot clear

# Start new focused conversation
gh copilot chat --new

# Include specific files in context
gh copilot ask "Explain this function" --include src/main.py

Switching Models Mid-Conversation

Temporary Switch

# Use different model for one question
gh copilot ask "Complex algorithm question" --model gpt-4

# Continue with default model
gh copilot ask "Simple follow-up"

Permanent Switch

# Change default model
gh copilot config set model gpt-3.5-turbo

# All subsequent commands use new model
gh copilot ask "This uses the new model"

Cost Optimization

Token Usage Monitoring

# Daily usage
gh copilot usage --today

# Weekly summary
gh copilot usage --week

# Model-specific usage
gh copilot usage --model gpt-4

Best Practices

  1. Start with GPT-3.5 for exploration
  2. Switch to GPT-4 for complex problems
  3. Use extended context models sparingly
  4. Clear context when switching topics
  5. Monitor usage regularly

Model Testing

Test Model Performance

# Test response quality
gh copilot ask "Explain recursion" --model gpt-4
gh copilot ask "Explain recursion" --model gpt-3.5-turbo

# Compare context handling
gh copilot ask "Summarize this large file" --include large-file.py --model gpt-4-32k

Benchmark Commands

# Time response speed
time gh copilot ask "Quick Python function" --model gpt-3.5-turbo
time gh copilot ask "Quick Python function" --model gpt-4

Configuration Validation

Check Current Model

gh copilot config get model

Verify Model Access

gh copilot model test gpt-4
gh copilot model test gpt-3.5-turbo

Reset to Defaults

gh copilot config reset model