Starting June 1, 2026, GitHub is overhauling how it bills Copilot users: flat monthly subscriptions are out, token-based pricing tied to actual AI consumption is in. The announcement has triggered one of GitHub's loudest public backlashes in years.
What's Changing
GitHub is moving all Copilot plans to a usage-based model. The new billing unit is called GitHub AI Credits, calculated from the tokens consumed in each interaction. Pricing varies by model — more capable models cost more credits per equivalent prompt.
Subscription prices stay the same on paper, but now they come with a base credit allowance. Once that's gone, extra usage is billed separately at token rates that vary by model.
Why Developers Are Furious
The official announcement thread collected nearly 900 downvotes and over 400 comments — an unusually strong reaction on GitHub's community forums. The core complaints:
- Unpredictable costs. Agentic sessions — refactoring, test generation, code review — can easily consume $30–40 in a single sitting, blowing through an entire month's Pro budget in a few hours.
- Same price, less value. Features previously covered by the flat subscription now incur separate charges.
- Meter shock. Developers used to working without usage limits now face a cost attached to every single operation.
GitHub's Reasoning
The company argues that Copilot has evolved far beyond simple code completion. Modern agentic workflows — where Copilot plans, executes, and iterates across multiple steps — consume vastly more compute, making flat pricing unsustainable at scale.
GitHub stresses that core features like code completions and Next Edit Suggestions remain within the subscription tier. Token billing kicks in mainly for chat interactions and agentic tasks.
What This Means Going Forward
Developers who rely heavily on Copilot for agentic tasks should review the new pricing structure and set spending caps in their account settings before June 1. GitHub has added usage monitoring tools, but many find them insufficient for real-time budget control.
The shift mirrors a broader industry move: as AI assistants grow more powerful, providers are switching from all-you-can-eat pricing to pay-as-you-go. Whether the developer community accepts this trade-off — or starts looking at alternatives — will be the story of the weeks ahead.
