GitHub will shift GitHub Copilot to usage-based billing starting June 1, charging users based on token consumption instead of a flat monthly request allocation.

GitHub announced a move from a flat monthly "requests" allocation to a usage-based model for GitHub Copilot, effective June 1. The company cited the need to "better align pricing with actual usage" and described the current system as "no longer sustainable" given escalating inference costs.

Under the new system, subscribers receive monthly "AI Credits" matching their subscription payment, with additional usage charged based on token consumption (input, output, and cached tokens) at API rates that vary by model—ranging from $4.50 per million output tokens (GPT-5.4 Mini) to $30 per million output tokens (GPT-5.5). Simple features like code completion and Next Edit remain free; code reviews will consume GitHub Actions minutes instead.

GitHub previously absorbed rising AI inference costs but can no longer do so uniformly, since identical request categories (e.g., a quick chat question versus a multi-hour autonomous coding session) incur vastly different backend computing costs.