Think Harder Mode: When to Use Advanced AI Analysis
What is Think Harder Mode?
Think Harder mode is an optional setting in AI Diff Review that uses more powerful AI models for deeper, more thorough analysis. When enabled, the plugin uses the most capable models available for your chosen provider, trading speed and cost for comprehensive insights.
Model Selection
When Think Harder mode is enabled, AI Diff Review automatically selects the most powerful model for your provider:
- OpenAI: Uses GPT-5 (instead of GPT-5-mini)
- Claude: Uses Claude Opus 4.1 (instead of Claude Sonnet 4)
- Gemini: Uses Gemini 2.5 Pro (instead of Gemini 2.5 Flash)
- Grok: Uses Grok-4-latest (instead of Grok-4-1-fast)
- Ollama/LM Studio: Uses your selected local model (no change, but you can choose a more powerful model)
When to Use Think Harder Mode
Complex Refactoring
When making significant architectural changes or large refactorings, Think Harder mode provides deeper analysis of the implications and potential issues. The more powerful models can better understand complex relationships and suggest improvements.
Security-Critical Code
For security-sensitive changes, Think Harder mode's deeper analysis can catch subtle vulnerabilities that might be missed in standard mode. The additional reasoning power helps identify complex attack vectors and security issues.
Performance-Critical Changes
When optimizing performance-critical code, Think Harder mode can provide more nuanced suggestions for improvements. The advanced models better understand performance implications and can suggest more sophisticated optimizations.
Unfamiliar Codebases
When working with code you don't know well, Think Harder mode can provide better explanations and context. The more powerful models are better at understanding code patterns and explaining what code does.
Important Commits
For commits that will be merged to main/master or are part of a release, Think Harder mode provides extra assurance that issues are caught. The additional analysis depth is worth the extra time and cost for critical changes.
When Standard Mode is Sufficient
Routine Changes
For everyday bug fixes, small features, and routine updates, standard mode provides excellent analysis without the extra cost and time. The standard models are quite capable for most use cases.
Quick Iterations
When you're iterating quickly and need fast feedback, standard mode's speed is more valuable than deeper analysis. You can always switch to Think Harder mode for the final review before committing.
Cost-Sensitive Workflows
If you're using cloud providers and are cost-conscious, standard mode provides good value. Reserve Think Harder mode for when you really need the extra depth.
Performance Considerations
Speed
Think Harder mode is slower than standard mode because:
- More powerful models take longer to process
- They may use longer timeouts
- They process more thoroughly
Expect 2-3x longer analysis times when using Think Harder mode.
Cost
For cloud providers, Think Harder mode costs more per request because:
- More powerful models have higher per-token costs
- They may use more tokens for thorough analysis
- Longer processing may incur additional charges
For local providers, there's no additional cost, but processing may take longer.
Configuring Think Harder Mode
You can enable Think Harder mode in two ways:
Global Setting
Go to Settings → Tools → AI Diff Review and enable "Think harder" model toggle. This makes Think Harder mode the default for all analyses.
Per-Analysis
Some analysis entry points allow you to enable Think Harder mode for that specific analysis. This gives you flexibility to use it only when needed.
Best Practices
Use Selectively
Don't enable Think Harder mode for every analysis. Use it strategically for important or complex changes where the extra depth is valuable.
Combine with Other Settings
Think Harder mode works best when combined with:
- Full content inclusion (gives models more context)
- RELAXED diff scope (comprehensive analysis)
- Higher related context caps (more surrounding code)
Review Timeouts
Think Harder mode enforces higher minimum timeouts. Make sure your timeout settings are appropriate for the longer processing times.
Monitor Costs
If using cloud providers, monitor your API usage when Think Harder mode is enabled. The additional cost can add up if used frequently.
Local vs Cloud Considerations
Local Providers
With Ollama or LM Studio, Think Harder mode doesn't cost extra, but you may want to use a more powerful local model. Consider using models like llama3, qwen2.5-coder, or other advanced options when Think Harder mode is enabled.
Cloud Providers
With cloud providers, Think Harder mode automatically uses the most powerful models, which cost more. Use it strategically to balance analysis quality with cost.
Conclusion
Think Harder mode is a powerful tool for getting deeper, more thorough code analysis. By using it strategically for complex, security-critical, or important changes, you can get maximum value while managing costs and time.
Standard mode is excellent for most use cases, so don't feel you need Think Harder mode for everything. Save it for when you really need the extra depth and are willing to trade speed and cost for comprehensive insights.
Experiment with both modes to understand when each is most valuable for your workflow. Over time, you'll develop a sense for when Think Harder mode is worth the extra investment.
Ready to try advanced analysis? Install AI Diff Review and explore Think Harder mode for your most important changes.