GitHub Copilot Drops Flat Pricing as Developers Face Steep Token Bills

GitHub Copilot is ending its flat-rate subscription pricing and switching to token-based billing starting June 1, 2026. The move by Microsoft's AI coding assistant has sparked immediate backlash from developers who say their monthly costs could skyrocket under the new consumption-based pricing model.

What Is Changing

Under the previous pricing structure, developers paid a fixed monthly fee — as low as 10 dollars for the Pro plan — and could use Copilot as much as they wanted without worrying about overages. The new billing system measures usage through tokens, which are the small chunks of text that large language models, or LLMs, process when reading and generating code. Each subscription tier now includes a set allotment of AI credits, and anyone who exceeds their limit will pay extra based on how many tokens they actually consume.

GitHub says the change is necessary because Copilot has evolved well beyond simple autocomplete. It now powers agentic workflows — automated multi-step coding tasks where AI handles everything from writing functions to reviewing pull requests. These complex operations consume significantly more computing power than basic code suggestions. The company stressed that simpler features like inline completions and Next Edit recommendations will remain included at no extra charge.

Developer Backlash Grows

Many developers have reported sticker shock after previewing their projected costs under the new model. One user shared that their monthly bill would leap from 29 dollars to roughly 750 dollars, while another estimated costs could surge from 50 dollars to approximately 3,000 dollars per month. Online reactions have been sharp, with some calling the new pricing “ridiculous” and announcing immediate plans to cancel their subscriptions.

Not everyone opposes the shift. Some developers argue that the extreme cost projections come from heavy reliance on fully automated “vibe coding” — sessions where AI handles entire projects with minimal human input, generating massive token volumes. They contend that teams using Copilot as a supplement rather than a replacement will still find the pricing reasonable.

The change reflects a growing pattern across the tech industry. Companies that once subsidized AI services with flat-rate plans are increasingly switching to usage-based pricing as the true infrastructure costs of running large language models become harder to absorb.