GitHub announced a change to usage-based billing for its Copilot AI service, effective June 1, to align costs with actual AI consumption. Subscribers will receive monthly AI Credits matching their subscription fees, with extra usage charged by token rates. The move addresses rising inference costs from heavy users.
GitHub, owned by Microsoft, revealed plans to replace its current request-based system for Copilot with one tied directly to AI resource use. Under the existing model, users get fixed monthly allocations of requests and premium requests, regardless of the varying computational demands of tasks like quick queries or extended coding sessions. GitHub stated that a simple chat and a multi-hour autonomous session currently cost users the same, a disparity it has subsidized until now but can no longer sustain amid escalating demand for AI resources. The new system provides each Copilot subscriber with AI Credits equivalent to their monthly payment. Beyond that, charges will apply based on tokens processed—input, output, and cached—using published API rates for models. For example, OpenAI's GPT variants range from $4.50 to $30 per million output tokens. Basic features such as code completion and Next Edit remain free of credits, while code reviews will incur costs via GitHub Actions minutes. Users can preview their bills under the new model before the June 1 rollout. This follows recent actions by GitHub, including pausing new subscriptions, tightening limits, and dropping advanced Claude Opus models from Pro plans to ensure service reliability. GitHub described the change as promoting sustainability by matching pricing to real usage and costs, reducing the need to restrict heavy users. Similar adjustments are underway elsewhere, with Anthropic charging enterprise users full compute costs and tweaking limits during peak hours.
