GitHub Copilot, the AI-powered coding assistant used by millions of developers worldwide, will transition from a flat-rate subscription model to a per-token billing system effective June 1, 2026. This change aligns the product’s pricing with how enterprise customers are typically charged for access to large language models (LLMs) through APIs.
Under the current model, users receive a fixed number of premium requests each month based on their subscription tier. Both complex multi-hour coding tasks and simple queries count equally as one premium request. The new system replaces this with token-based measurement, where most requests are charged according to the number of tokens processed by the underlying LLM.
How Tokens Are Defined and Billed
A token represents approximately three-quarters of a word in natural language. For example, a 10,000-word text input would equate to roughly 12,000 to 13,000 tokens. In coding contexts, this includes expressions, statements, variable names, functions, and similar elements. Prompt text, inputs, and outputs from Copilot all count toward token usage.
GitHub is keeping its pricing tiers at current levels but replacing the monthly query allowance with AI Credits. A Copilot Pro subscriber paying $10 per month will receive 1,000 credits, with GitHub currently valuing each credit at one US cent. The number of tokens per credit depends on the model used, the ratio of input to output, cache size (data retained in the LLM’s memory for context), and the specific feature requested.
Developers using mostly simple queries may find they do not need to purchase extra credits each month. Conversely, multi-agent queries involving complex, lengthy code bases will deplete the AI Credit account more quickly. Queries directed to the most advanced frontier models will incur higher costs than those using less powerful alternatives.
Compensatory Features and Industry Context
GitHub has included some compensatory benefits: code completions (similar to auto-complete functions) and next edit suggestions will remain free of charge. The change mirrors similar moves by other AI companies. Anthropic and OpenAI have already shifted their enterprise customers to token-based billing.
Unlike those companies, Microsoft, which owns GitHub, is a profitable enterprise overall. Up to now, Microsoft has been able to subsidize Copilot usage with revenues from its software and cloud divisions. Until June 1, users effectively spent between three and eight times the token value of their monthly subscription without penalty.
Implications for Developers and Businesses
Microsoft’s decision affects the very developers it hoped to attract to Copilot. New and existing users must now become aware of their token spend per query, a figure that was previously abstracted by flat monthly subscriptions. While the new billing model may make economic sense from Microsoft’s perspective, it may discourage the exploration and testing that new users typically want to perform.
For businesses deploying AI coding agents in development teams, the cost implications are substantial. Uber’s chief technology officer, for example, told The Information that the company had already exhausted its 2026 AI budget earlier this year, noting that 11 percent of updates to Uber’s code are now written by AI, primarily using Anthropic’s Claude coding agents.
Broader Impact Beyond IT Departments
Companies deploying AI automation outside IT should also take note. Complex tasks that involve running agentic LLMs unsupervised for extended periods may soon be charged on a similar per-token basis. Any efficiency gains from AI in the workforce will need to be measured against potential increases in AI vendor bills.
Industry observers expect other AI coding tool providers to adopt similar pricing models in the coming months. As token-based billing becomes the standard, organizations will need to develop strategies for monitoring and optimizing their AI usage to control costs.
