GitHub’s AI code assistant, Copilot, adopted a token-based billing model on June 1st, marking a significant change from its previous fixed-fee premium model. The new model introduces “AI Credits,” where developers will pay based on the number of tokens consumed during advanced interactions, while the base subscription price remains unchanged. GitHub Chief Product Officer Mario Rodriguez announced the shift on April 27th, citing the increasing computational demands of Copilot’s agent functionality as the primary reason for the change.
This move has sparked widespread concern among developers. Discussion threads on the GitHub community regarding the change have accumulated over 400 comments and nearly 900 dissenting votes, with many developers warning that users who frequently use the agent coding functionality could face significantly increased costs. One developer estimated that a single session using Copilot’s agent functionality could consume $30 to $40 worth of credits, three to four times the monthly allowance for a Pro member.
In the new system, 1 AI Credit is equal to $0.01, and each interaction is metered based on input, output, and cached tokens. The cost depends on the selected AI model, the length of the prompt, and the complexity of the output. Credits do not carry over to the next month.
GitHub is replacing its Advanced Request Unit system with AI Credits, which applies to all four Copilot plans. The $10/month Copilot Pro includes $10 in AI Credits, while the $39/month Copilot Pro+ includes $39. Business and Enterprise subscriptions are priced per user and offer temporary promotional credits of $30 and $70 respectively, valid until August 2026, to facilitate the transition. Annual subscribers will continue to use the old billing model until their current subscription expires.
Importantly, Copilot’s two core features—code completion and “next edit suggestions”—remain unlimited and unaffected by the new token-based billing method. The metered mode applies only to chat interactions, smart workflows, and code review tools.
GitHub stated that this move is a necessary adjustment to reflect real-world usage. Previously, when usage exceeded a threshold, GitHub would revert to a low-cost mode, but this practice was criticized by developers for reducing quality. The new system allows users to choose their preferred AI models and pay corresponding token fees. To help enterprise customers, GitHub introduced multi-level budget control and launched a billing preview feature in May so users can understand the expected costs.