Right-Size Your PR Reviews: GitHub Copilot Effort Levels (Lite vs. Balanced) Are Now GA

GitHub Copilot's code review now has two depth modes — Lite for straightforward changes and Balanced for complex ones — so review compute matches the actual risk of each pull request, not a fixed ceiling.

GitHub shipped Lite and Balanced code review effort levels to general availability on August 7. Here's what each does and how to configure them.

The problem they solve

Before effort levels, every PR triggered the same Copilot review depth — overkill for a documentation fix, and sometimes insufficient for a cross-service auth refactor. The levels let you express "routine vs. complex" intent and get a calibrated response.

Lite vs. Balanced

Lite uses a lighter, faster model. Reach for it on:

• Documentation or copy fixes
• Config-only changes
• Formatting adjustments
• Narrow, well-isolated bug fixes

Balanced uses a higher-reasoning model for deeper analysis. Use it on:

• Security-sensitive changes (auth, permissions, cryptography)
• Cross-service or cross-API refactors
• New external-facing contracts or APIs
• Changes touching shared infrastructure or data models

Both levels work across Copilot Pro, Pro+, Max, Business, and Enterprise.

Configuration

Per-review: Select the level when requesting a review. This overrides the org default for that one PR only — useful when a routine PR suddenly touches sensitive code.

Org-wide default: Go to Organization Settings → Copilot → Copilot code review and set the default. All repositories inherit it. Individual repos or reviewers can override it per-review.

Audit trail: The effort level now appears in PR timeline events and inline review comments, so you can always see which depth was applied to any historical review.

Practical setup

A sensible starting configuration:

• Set org default to Lite — keeps baseline cost and latency low
• For repositories flagged as security-sensitive (auth service, billing, infra), configure Balanced as the repo-level default in repository settings
• Engineers can always request Balanced per-review on any PR that warrants it

The key insight: Copilot reviews have cost in two directions — compute and time. Defaulting to maximum depth universally wastes both; defaulting to minimum depth misses real issues. Effort levels give you the control to match depth to risk without per-PR manual triage.

Sources: Copilot code review effort levels GA — GitHub Changelog · Configuring automatic code review — GitHub Docs