GitHub Copilot's Flat $10 Plan Quietly Became Pay-Per-Token in June. Developers Are Still Finding Out the Hard Way
For two years, Copilot was simple: pay a flat monthly fee, get Copilot. GitHub's April 27, 2026 announcement ended that. Code completions and Next Edit suggestions stayed free and unmetered. Everything else, chat, agent mode, code review, draws down a credit balance priced off the token cost of whichever model is actually doing the work.
The quiet part of the change is what it removed: when a plan used to run out of its premium-request allotment, Copilot fell back to a cheaper model automatically. That fallback is gone. Run out of credits now and usage is governed directly by your remaining balance and any admin-set budget, full stop.
What GitHub actually changed
| Plan | Seat price | Stays unmetered | Now draws credits |
|---|---|---|---|
| Pro | $10/month | Code completions, Next Edit | Chat, agent mode, code review |
| Pro+ | $39/month | Code completions, Next Edit | Chat, agent mode, code review |
| Business | $19/user/month | Code completions, Next Edit | Chat, agent mode, code review |
| Enterprise | $39/user/month | Code completions, Next Edit | Chat, agent mode, code review |
Credits are calculated from input, output, and cached tokens at each model's published API price, per GitHub's announcement. Code review additionally consumes GitHub Actions minutes, billed separately.
Why heavy users are the ones who actually feel it
A single chat question and a multi-step agent-mode session that reads a codebase, edits several files, and re-checks its own work consume wildly different amounts of tokens, and therefore wildly different amounts of credit, even though both used to just count as "Copilot usage" under the old flat plan.
June 1, 2026
the day flat-rate Copilot billing ended on every paid plan
⅔ of a month
of credits one developer says they burned through in a single day of light agent-mode work
reported to Visual Studio Magazine, as of August 7, 2026.
1,000s
of replies piled onto GitHub's own Discussions thread about the change within its first two weeks
Keeping the bill predictable
- 1Set a per-user spending cap in admin billing settings before rolling Copilot out to a whole team, rather than after the first surprise invoice.
- 2Default to a cheaper model for everyday chat and reserve the frontier model for agent-mode tasks that genuinely need it; the credit cost difference between models is the whole lever here.
- 3Watch the credit dashboard GitHub added after the backlash, not just the monthly invoice. Allotments reset monthly and don't roll over.
- 4If your team's work is mostly agentic coding, run the numbers against alternatives like Cursor, Windsurf, or Anthropic's Claude Code before assuming Copilot is still the cheaper option; which one wins now depends entirely on your actual usage pattern, not the sticker price.
Did GitHub Copilot's price go up?
Seat prices didn't change: Pro is still $10/month, Pro+ $39/month, Business $19/user/month, Enterprise $39/user/month. What changed is that AI usage beyond code completions now draws down token-metered credits instead of a flat monthly allotment.
When did GitHub Copilot move to usage-based billing?
GitHub announced the change on April 27, 2026, and it took effect for all paid plans on June 1, 2026.
What still doesn't use credits?
Code completions and Next Edit suggestions remain free and unmetered on every paid plan. Chat, agent mode, and code review all draw from your credit balance.
What happens when I run out of credits?
There's no more automatic fallback to a cheaper model. Usage is governed directly by your remaining credit balance and any spending limit an admin has set.
Why are some developers' bills so much higher now?
Agent-mode sessions that read and edit code across multiple files consume far more tokens, and therefore far more credit, than simple chat questions or completions did under the old flat-rate system.
Try the tools ✦
Free browser tools that never upload your files.