GitHub Copilot will change its billing model on June 1, 2026. The company will replace today’s premium request units with GitHub AI Credits, a usage-based system that will apply to all Copilot plans. Base subscription prices are not changing, but the way usage is measured is: each advanced AI interaction will consume credits based on input, output and cached tokens.
The move marks an important shift in the programming assistant market. Copilot started as an in-editor helper designed to complete code, suggest functions and answer quick questions. It is now becoming an agentic platform capable of running long sessions, reviewing entire repositories, working on pull requests and using increasingly expensive models. That technical leap has a direct consequence: the flat-rate model is starting to show its limits.
From premium requests to GitHub AI Credits
Until now, GitHub Copilot used a premium request system to measure part of the usage of advanced models. From June 1, that structure will be replaced by monthly credits included in each plan. The economic unit will be more transparent: 1 AI Credit equals $0.01. The cost of each interaction will depend on the model used and the number of tokens processed.
GitHub says input tokens, output tokens and cached tokens will all be counted. This means a quick question will not cost the same as a long agentic session in which Copilot reads repository context, generates changes, iterates on errors and produces a lengthy response. The company argues that the previous model no longer reflected the cost difference between these two types of use.
Base prices remain unchanged. Copilot Pro will still cost $10 per month, Pro+ $39, Business $19 per user per month and Enterprise $39 per user per month. The difference is that those amounts will become monthly included credits, with the option to buy additional usage on paid plans if the user or organisation allows it.
| Plan | Monthly base price | Credits included from June |
|---|---|---|
| Copilot Pro | $10 | $10 in AI Credits |
| Copilot Pro+ | $39 | $39 in AI Credits |
| Copilot Business | $19/user | $19 in AI Credits |
| Copilot Enterprise | $39/user | $39 in AI Credits |
During the transition, GitHub will give businesses and organisations more room to adjust. Existing Copilot Business customers will receive $30 in monthly included credits during June, July and August. Copilot Enterprise customers will receive $70 in monthly included credits during the same three months.
Completions remain included, agents do not
GitHub has tried to separate two types of usage. Code completions and Next Edit suggestions will remain included in all plans and will not consume AI Credits. This is an important decision because it protects Copilot’s most common editor-based use: autocompleting lines, continuing a function or suggesting the next change.
Usage will be concentrated in more intensive capabilities, especially chat, agents, code reviews and long sessions. That is where the real change happens. GitHub acknowledges that Copilot is no longer the same product it was a year ago. The company describes its evolution into an agentic platform capable of executing multi-step tasks and working across entire repositories.
Fallback experiences will also disappear. Until now, when users exhausted premium requests, they could fall back to a lower-cost model and continue working. Under the new model, usage will be governed by available credits and budget controls set by administrators.
Copilot code review will also have a specific cost structure: it will consume AI Credits and GitHub Actions minutes when it runs on private repositories. GitHub says this is because agentic reviews rely on tool-calling architecture and GitHub Actions runners to analyse repository context and produce more relevant pull request comments.
What changes for companies and engineering teams
For individual users, the impact will depend on how they use Copilot. Those who mainly rely on completions and lightweight queries may notice little change. Those who use agents, automated reviews, advanced models and long sessions will need to monitor consumption more closely. Monthly Pro and Pro+ users will migrate automatically on June 1. Annual plan users will remain on premium request-based pricing until their subscription expires, although model multipliers will change from June.
For companies, the change has broader implications. GitHub will introduce pooled included usage across organisations, helping avoid wasted credits from low-usage users while others exceed their allocation. Administrators will also be able to set budgets at enterprise, cost centre and user levels. Once the included pool is exhausted, the organisation can decide whether to allow additional usage at published rates or cap spending.
That control will be necessary. In development teams, AI consumption can vary significantly between profiles. A developer who uses Copilot mainly for autocompletion does not generate the same load as someone running agentic sessions to resolve issues, review pull requests, generate tests or explore legacy code. The new model puts a price on that difference.
The message for engineering and finance leaders is clear: AI-assisted development is entering a FinOps phase. It will no longer be enough to count active licences. Teams will need to measure real usage, cost per team, cost per repository, productivity impact and the return on agentic features. Copilot will remain useful, but heavy usage will leave a much more visible financial signal.
The end of the illusion of unlimited AI
GitHub’s change is not an isolated case. The industry is moving toward usage-based billing because generative AI, especially when it works as an agent, consumes resources in highly variable ways. A short conversation and a multi-hour task do not have the same inference cost. In the first stage of the market, many providers absorbed that difference to accelerate adoption. That phase is starting to end.
The new model may also change working habits. Teams will need to decide when it is worth using an advanced model, when an included completion is enough and when to limit long agentic runs. A new internal discipline may emerge: optimising prompts, reducing unnecessary context, avoiding pointless retries and designing more efficient workflows.
The risk is that some organisations react with controls that are too strict and slow down valuable use cases. The opposite risk is also real: leaving consumption open and discovering too late that agents have generated a bill that is hard to justify. GitHub is trying to soften the transition with preview billing tools from early May, usage reports and budget controls.
For developers, the message is mixed. Copilot will continue to include completions and keep base prices unchanged, but advanced usage will have clearer economic limits. For GitHub, the change is about sustaining a product that has grown from an editor assistant into an automation layer across the software development lifecycle.
The conclusion is simple: AI for programming is no longer billed only by seat. It is starting to be billed by work performed, measured in tokens, models and infrastructure minutes. Copilot is not leaving flat-rate pricing entirely, but flat-rate pricing will no longer cover everything.
Frequently asked questions
When does GitHub Copilot billing change?
The new GitHub AI Credits model will take effect on June 1, 2026 for all Copilot plans.
Is Copilot’s base price increasing?
No. Copilot Pro will remain $10 per month, Pro+ $39, Business $19 per user per month and Enterprise $39 per user per month.
Which features will remain included without consuming credits?
Code completions and Next Edit suggestions will remain included in all plans and will not consume GitHub AI Credits.
What should companies do before the change?
They should review the preview billing experience, download usage reports, define budgets by organisation, cost centre or user, and decide whether to allow additional consumption once included credits are exhausted.