From June 1, 2026, GitHub Copilot plans no longer count premium requests. Instead, every plan receives a monthly allocation of GitHub AI Credits, with usage above that allocation billed at token rates derived from the underlying model’s API pricing.
The new structure: Pro at $10/month includes 1,500 credits, Pro+ at $39 includes 7,000 credits, and a new Max tier at $100/month includes 20,000 credits. Cached tokens cost less than output tokens. Input tokens cost less than output tokens. The specific rate depends on which model is being invoked.
Why GitHub made this change
The fixed premium request model was a pricing layer built on top of underlying token costs — GitHub absorbing the difference between what users consumed and what they paid. As the models got more capable (and more expensive per interaction), that subsidy became harder to sustain. Token billing passes the actual cost of inference through to the user, more or less directly.
It also reflects a technical reality: not all Copilot interactions cost the same. A multi-file refactoring task that invokes a high-capability model with a long context window costs substantially more than a single-line autocomplete. Fixed request counts couldn’t reflect that variance. Credits can.
What it means for development costs
For individual developers with light usage, the credit allocations are probably generous enough that the change won’t matter. For teams doing heavy AI-assisted development — extended agentic sessions, large codebase analysis, high-frequency completions throughout the day — it’s a meaningful shift.
The comparison relevant to professional development teams: Cursor restructured its Teams pricing in June 2026 to $32/seat/month on Standard and $96/seat/month on Premium. Claude Code is available at $100/month for individual Pro plans. GitHub Copilot Business is $19/user/month. The absolute numbers are becoming less meaningful than understanding what’s included and at what cost above the threshold.
The broader signal
GitHub is not an outlier here. The move toward token-based consumption pricing is happening across the AI tooling market, and it reflects the economics of the underlying infrastructure. The era of flat-rate AI tools with unlimited usage is ending, roughly in proportion to how useful those tools have become — the more they’re used in anger on real projects, the less sustainable fixed pricing becomes.
For founders and product leads commissioning software, this is worth factoring into total project costs. AI tooling has moved from a discretionary add-on to a core line item in how professional development teams operate. The efficiency gains are real; so is the cost of maintaining them.
See our AI-assisted development page for more on how we use these tools in practice.