Model Selection

Core Features
3 min read

Model Selection

ThinkReview automatically picks the best AI model for your code reviews. You don't need to worry about it—it just works.

How it works

When you request a review, we look at several factors:

  • Your subscription tier — Priority access based on Free, Lite, Pro, or Teams tiers.

  • Your preferences — Models you've "Selected" in settings are tried first.

  • Your code size — We match the patch length to the model's context window.

  • Exclusions — Models marked as "Restrict from Use" are completely blocked.

    • Selection Logic

    ThinkReview uses a priority-based selection algorithm to give you the best review possible:

  • Tier Priority: Models are categorized by tier (Teams, Professional, Lite, Free). Higher tiers have access to more sophisticated models with larger context windows.

  • Preferred Selection: If you have selected specific models, the system will attempt to use those first, provided they are compatible with your current tier and the size of your code.

  • Smart Truncation: For large patches, our smart truncation engine ensures we process the most critical parts of your code while staying within model limits.

    • Fallback Strategy (Backup Plan)

    Sometimes a model might be slow or return an error. In these cases, ThinkReview doesn't stop. We automatically attempt up to 3 fallback attempts:

    • Intelligent Rerouting: If a model fails, we prioritize the next best available model.

    • Size-Based Priority: We often fallback to models with larger context windows than the one that failed to minimize truncation.

    • Reliability: If premium models are unavailable, we fall back to highly stable alternatives within your tier.

      • Pro tip: You can customize which models to use in your settings.

      Customize it

      Want to pick your own models? Head to Model Selection settings and choose:

    • Select Models — Pick which models you want to use for reviews (your preferred models).

    • Restrict from Use — Block specific models from being used (both for initial selection and fallback).

      • Selecting Preferred Models

      Choose which models you want ThinkReview to prioritize for your code reviews. The system will try these models first before falling back to others within your tier.

      Restricting Models

      If there are models you don't want to use, you can restrict them. Restricted models will be completely blocked from:

    • Initial model selection

    • Fallback attempts (if your primary model fails)

      • ⚠️ Important: You must have at least 2 models that are not restricted. This ensures the system has backup options in case any models produce an error.

      Troubleshooting

      Reviews taking longer than usual?

    • Large patches require more analysis time. Try smaller commits for faster feedback.

    • If your preferred models are heavy "Reasoning Models", they may take longer to process.

      • Not happy with review quality?

    • Try a different model in settings.

    • Restrict models that don't match your coding style.

    • Use the Refresh button for a second opinion.

      • Getting errors about restricted models?

    • Make sure at least 2 models are not restricted. This is a system safety requirement.

    TL;DR: We automatically pick the best AI model based on your subscription, preferences, and code size. It just works. Customize it if you want, but the defaults are solid.