GitHub starts Copilot Max plan June 1

GitHub is changing what “paying for Copilot” means. Starting June 1, individual plans stop being mostly flat subscriptions with soft limits and start looking more like cloud services — you get included usage, then you watch the meter. The headline change is a new Copilot Max plan for people who live in Copilot all day, plus bigger built-in “flex” allotments for Pro and Pro+ users. But the real story is the billing model under it. ### What changed on June 1? GitHub said the individual lineup will be Free, Pro, Pro+, and Max, with usage-based billing across all plans beginning June 1, 2026. Free still gets limited completions, chat, and agent usage. Paid plans now bundle monthly included usage in two buckets — base credits tied to the subscription and extra flex allotments for Pro and Pro+ — while Max is the new option for sustained high-volume use. (github.blog) ### What is Copilot Max, really? Basically, it is GitHub admitting that one kind of Copilot user broke the old pricing model. Casual autocomplete users and power users running long chats, agents, and code review loops do not consume remotely the same amount of compute. Max is the tier for that second group — the people using premium models and agentic workflows often enough that Pro or Pro+ would feel cramped even with the new flex buffer. GitHub framed it as a plan for “sustained, high-volume” work. (github.blog) ### What are “AI Credits”? They are the new billing unit underneath Copilot. Instead of counting premium requests, GitHub is moving all Copilot plans to GitHub AI Credits, with usage calculated from token consumption — input, output, and cached tokens — using model-specific API rates. That is a big shift because it makes cost depend more directly on how much model work you trigger, not just whether you crossed a feature gate. (github.blog) ### Why did GitHub add flex allotments? Because the company already tried the blunt version first. In April, GitHub paused new self-serve signups for some individual plans, tightened limits, and adjusted model availability to protect service quality for existing customers. Flex allotments look like the cleaner follow-up — more breathing room for paid users without pretending all usage is equal. It is a pressure valve, but one that still keeps the meter intact. (github.blog) ### Why are April reports a big deal? Because GitHub is giving customers a dress rehearsal before the meter goes live. As of May 12, Copilot Business and Enterprise admins can download April usage reports showing how that activity translates into AI credits ahead of June 1 billing. That turns an abstract pricing change into a spreadsheet problem — who used what, where, and how much it would have cost. (github.blog) ### What about code review? GitHub tweaked that too. Copilot code review comments now show grouped suggestions, severity levels, and an updated suggested changeset UI for users on the new pull request experience. Separately, GitHub added code review comment types to the usage metrics API, and it previously said Copilot code review will start consuming GitHub Actions minutes on June 1. So review is getting easier to scan, but also easier to price. (github.blog) ### Who actually has to care? Individual power users first, but teams right behind them. Contractors, startups, and engineering managers now have to forecast Copilot consumption the way they already forecast cloud spend. The catch is that AI usage is spikier than seat licenses — one engineer doing lots of agent work can consume far more than another with the same plan. That makes procurement, chargebacks, and rollout decisions more complicated. This is partly inference, but it follows directly from GitHub moving to token-based credits and exposing usage reports before billing starts. (github.blog) ### So what is the bottom line? Copilot is becoming a metered developer utility. GitHub is still selling subscriptions, but the center of gravity is shifting from “which plan are you on?” to “how much model work did you actually do?” Max is the clearest sign yet that heavy AI coding use is no longer an edge case — it is its own product category. (github.blog 1) (github.blog 2)