GitHub Copilot's new usage-based pricing model took effect today, and power users are reporting severe sticker shock. The shift from request-based to credit-based billing has some developers burning through their entire monthly allotment in a single day.
Under the new system, Copilot subscriptions grant a set number of AI 'credits' each month-one credit equals $0.01 of usage. The $10/month Pro plan includes 1,500 credits ($15 worth), while the $39 Pro+ plan offers 7,000 credits ($70). The $100/month Copilot Max plan includes 20,000 credits ($200).
Costs vary by model. A simple chat query might consume 15 credits, while a complex prompt can burn through 171 credits or more. One user reported spending 840 credits 'being super cautious.'
GitHub had previously announced the change in April, arguing that the old system forced the company to 'absorb much of the escalating inference cost.' Some users, looking at GitHub's own cost calculation tools, say their previous monthly usage would now rack up bills in the thousands of dollars.
In response, many are threatening to cancel subscriptions or seek alternative AI coding tools. But some developers report adapting by making 'very focused and deliberate changes' and avoiding long-running chat sessions that send entire histories as context.
Industry observers note that if other AI coding services follow GitHub's lead to usage-based pricing, the most token-efficient LLMs-like Deepseek-may gain a competitive advantage.