What AI Code Review Costs After the GPT-5.6 Luna Price Cut

This is a real review from Cody's Private AI PR Reviewer on a public pull request, produced by GPT-5.6 Luna. The header names the model.

How it works

Cody is a GitHub Action. A pull request opens, the workflow reads the diff through the GitHub API, sends it with a focused review prompt to the model on your own API key, and posts the review as a pull-request comment. There is no subscription and the unmodified Action has no integration with a Cody Labs-operated data service; you pay your selected model provider for tokens, and GitHub Actions minutes may apply on private repositories.

What it cost

GPT-5.6 Luna is now US$0.20 per million input tokens and US$1.20 per million output tokens. The review above sent roughly 2,300 tokens of diff and prompt and returned a 2,370-character review. In visible tokens that is about a tenth of a cent.

The account's own meter is the simpler argument. Every request this billing window, including that review, totals 6,244 tokens across 15 requests, and the spend reads $0.00:

Token figures are estimates at roughly four characters per token and exclude any hidden reasoning tokens; the dashboard total is the real bill.

What that adds up to

At these sizes a dollar covers hundreds of reviews. A team merging 100 pull requests a month would spend well under a dollar on review inference. The model cost of a review is now small against the human time it saves.

Using GitLab or Azure DevOps? Cody Pro runs the same reviews in GitLab CI and Azure DevOps pipelines.