GitHub will start charging Copilot users based on their actual AI usage

In a significant update, GitHub has announced plans to implement a usage-based billing model for its Copilot AI service, set to take effect on June 1. This change aims to align costs more closely with actual user engagement, addressing the growing demand for limited AI computational resources. Currently, GitHub Copilot users benefit from a monthly allocation of standard and premium requests, which are utilized whenever they seek assistance from the AI model. However, GitHub acknowledges that these categories encompass a wide array of AI tasks that incur varying backend costs. As stated in their announcement, "Today, a quick chat question and a multi-hour autonomous coding session can cost the user the same amount." While GitHub has previously covered a significant portion of the rising inference costs, the company has determined that the previous method of categorizing all premium requests together is no longer viable. Under the new system, subscribers will receive a monthly supply of "AI Credits" that corresponds to their subscription fee. Users who exceed these credits will incur charges based on token consumption, which includes input, output, and cached tokens, with rates varying by the model’s complexity. For instance, the cost for OpenAI's advanced GPT models ranges from $4.50 to $30 per million output tokens. The token count for a specific AI prompt can fluctuate significantly based on the complexity of the task at hand. Importantly, users will still have access to basic features like code completion and Next Edit without depleting their AI credits. However, engaging in Copilot code reviews will incur additional charges in the form of GitHub Actions minutes.