Microsoft's GitHub officially announced on April 28, 2026, that its AI coding tool GitHub Copilot will transition from a fixed subscription model to a usage-based billing model starting June 1. This change introduces a new "GitHub AI Credits" credit system, aimed at addressing the increased computing power demands brought by Copilot's evolution from a single assistant to a multi-step autonomous coding intelligent platform.

Github

According to the new rules, the original "Premium Request Unit" (PRU) counting method for GitHub Copilot will be abolished. The base subscription prices of each plan remain unchanged, but instead, a monthly allocation of equivalent AI Credits will be provided: Copilot Pro (10 dollars per month), Pro+ (39 dollars per month), Business (19 dollars per person per month), and Enterprise (39 dollars per person per month) will receive corresponding credit values. Usage will be billed based on token consumption (including input, output, and cache) according to the model's public API rates. Notably, basic code completion and Next Edit suggestions will still not consume credits.

For enterprise customers, the new model provides a more flexible "organization shared credit pool," allowing light users' unused credits to offset heavy users' consumption. Administrators can set budget limits across different levels. To ensure a smooth transition, GitHub will offer enterprise users an additional promotional credit allowance of 30-70 dollars for three months. For existing annual subscribers, the original terms will remain in effect until the end of their subscription period, but starting June 1, the model multiplier will significantly increase, and after expiration, they will need to switch to a monthly pay-as-you-go model.

This billing transformation reflects the inevitable trend in the AI industry moving from "all-in-one" monthly packages to "precision pricing." By adopting a usage-based billing model, GitHub aims to maintain long-term service reliability while breaking through credit limitations in heavy usage scenarios, providing commercial support for developers to access higher-performance large-scale models.