GitHub Copilot shifts to a per-token pricing model starting June 1, 2026. The company abandons its flat-rate subscription approach, which gave users a fixed number of "Premium Requests" each billing cycle. The new system charges users based on actual token consumption, aligning Copilot's pricing with how most large language model APIs operate.
This change affects how developers budget for AI coding assistance. Users who consume fewer tokens will pay less, while heavy users face higher bills. The previous subscription model offered predictability. The token-based approach introduces variability but rewards conservative usage patterns.
GitHub's decision reflects industry trends toward usage-based pricing for AI services. Companies like OpenAI and Anthropic use similar token counting mechanisms. The transition requires users to monitor their consumption more closely.
The timing gives users several months to adjust their workflows and understand token economics before the change takes effect. Developers will need to track how many tokens their coding sessions consume to estimate future costs.