GitHub Copilot kills premium requests on June 1. Token billing arrives, fallback models do not.

GitHub will retire premium request units on June 1, 2026 and charge Copilot users by the token instead. Plan prices stay flat. Pro is still $10 a month, Pro+ $39, Business $19 per seat, Enterprise $39, with each tier swapping its old request quota for an equal dollar amount of GitHub AI Credits.

That math doesn’t change much for people who only run chat-style autocomplete. It changes a lot for anyone whose Copilot session is running an agent on Claude Opus or GPT-5.5 for hours. The April 27 announcement from GitHub Chief Product Officer Mario Rodriguez frames the shift as a sustainability move; the small print is harsher for annual subscribers, who don’t get to opt in.

What’s actually changing

GitHub AI Credits replace premium request units (PRUs) on June 1. Credits get burned at the published API rates per model, counting input, output, and cached tokens. Plan prices don’t move: Pro keeps $10 a month with $10 of credits, Pro+ stays $39 / $39, Business $19 per user / $19, Enterprise $39 / $39.

Code completions and Next Edit suggestions stay free across every plan and don’t touch the credit pool. The piece that goes away is the fallback experience. When credits run out, Copilot stops. It no longer downgrades the session to a cheaper model so the seat keeps working. Copilot code review now also consumes GitHub Actions minutes on top of credits, billed at the standard Actions per-minute rate.

Business and Enterprise customers get a three-month promotional bump: $30 of monthly AI Credits per Business seat and $70 per Enterprise seat across June, July, and August. Pro and Pro+ get nothing equivalent.

Why GitHub did it now

“Copilot is not the same product it was a year ago,” Rodriguez wrote in the announcement post. “Today, a quick chat question and a multi-hour autonomous coding session can cost the user the same amount.” He went on to call the existing PRU model “no longer sustainable.” The April 20 Pro reshuffle that booted Claude Opus to the $39 Pro+ tier was, in GitHub’s own words, a “reliability and performance” patch ahead of this transition.

The cost pressure is real. Ed Zitron at Where’s Your Ed At reported the week-over-week cost of running Copilot has nearly doubled since January, driven by agent workloads chewing through Opus tokens at $5 input and $25 output per million. Microsoft has been quietly absorbing that gap. The new model hands it to the customer.

What’s still up in the air

The community discussion thread that opened on April 27 surfaces what GitHub hasn’t pinned down.

• Whether AI Credits will ever roll over month to month. As written, they don’t, so seasonal heavy users subsidize their own light months.
• The full annual-plan multiplier table. Annual subscribers stay on PRUs until renewal, but the multipliers go up June 1, and the docs have changed weekly.
• Whether Pro and Pro+ get any version of the Business and Enterprise three-month promo credits. Right now: nothing.
• How hard the fallback removal hits the Free tier. Paid plans at least have a budget; the Free experience could thin out.
• Refund cutoffs after May 20. After that date there is no opt-out path back to PRU pricing.

What this means for you

If you use Copilot for autocomplete and the occasional chat, June 1 won’t show up on your card. If your team runs Codex-style agents through Claude Opus 4.7 or GPT-5.5 inside Copilot, plan a budget audit before May 31. The preview bill that GitHub is rolling out in early May will tell you what the same workflow costs in tokens; expect surprises on long agent runs. Annual subscribers should compare a prorated migration to a monthly plan against riding through renewal at higher multipliers. And if you’ve been weighing a Cursor subscription or running Claude Code directly through Anthropic, the gap between metered Copilot and a flat-rate competitor just narrowed.