Model Inversion

A different model reviews the code. Pull requests written by Claude are reviewed by GPT, and those written by Codex are reviewed by Claude, to improve review accuracy.

Book a demo All use cases

The opposing model reviews the PR.

If a PR was generated by Codex or Claude, automatically use the opposing model for review to improve accuracy.

GitHubpayments-api #318 / reviewer selection

Commit message

feat(payments): make retry cap configurable

Read RETRY_CAP from config, default 3.

Co-authored-by: Claude <noreply@anthropic.com>
Detected author
Claude
Reviewed by
GPT reviewer family

The configured GPT reviewer gives Claude-written code a second model's perspective.

A second opinion
from another model.

Code written by one model family is reviewed by the other, so every AI-written pull request gets a second model's perspective. It helps find more bugs and issues.

Review AI code with the other model.

Code attributed to Claude is reviewed by GPT. Code attributed to Codex is reviewed by Claude.

  • Written by ClaudeGPT reviews
  • Written by CodexClaude reviews
  • Not recognizedYour usual model
Try the example

Detected from commit trailers.

CloudAEye reads recognized Co-authored-by trailers in commit messages to tell which model wrote the code.

  • ClaudeCo-authored-by: Claude
  • CodexCo-authored-by: Codex
  • No trailerUsual model stays
Read the docs

On pull requests and before commit.

Inversion applies to pull request reviews, and to pre-commit reviews in Claude Code, where it selects the configured GPT reviewer.

  • Pull requestsGitHub, GitLab, Bitbucket
  • Claude Code/review, /inspect, /security
  • Before commitGPT reviewer selected
Turn it on

Override any single run.

Pick the reviewer for one review with a flag. An explicit model always wins, and flags never change your saved settings.

  • --model, --model-providerOverrides inversion
  • --invert-model true|falseClaude Code commands
  • Saved settingsUnchanged by flags
See the commands

Set it once for your team.

Turn it on

  1. Open Code Review Settings > LLM.
  2. Configure and save your preferred Code Review model.
  3. Turn on Model inversion.

Tenant owners and admins can change this setting. Available on Pro and Enterprise plans. Turning it off restores your configured model.

How the reviewer is chosen

  1. An explicit model wins. --model or --model-provider overrides inversion for that run.
  2. Otherwise, the run's inversion flag. --invert-model true or false in Claude Code review commands.
  3. Otherwise, your saved setting. Model inversion on or off for your organization.

Overrides apply to one run only and never update your saved configuration.

Commands

Claude Code
/cloudaeye:inspect --invert-model true
GitHub · Bitbucket · GitLab
@cloudaeye /inspect --model-provider openai --model gpt-6.1-sol

With inversion on, the reviewer is picked automatically. Use these to choose the reviewer for a single run.