Two AI Coding Tools Reset Their Pricing This Weekend. Copilot Goes Usage-Based Tomorrow and Codex Pro's 10x Promo Dies Today.
GitHub Copilot moves every plan to usage-based billing on June 1, 2026, that's tomorrow. The same weekend, OpenAI's $100/month Codex Pro tier quietly drops from its launch-promo 10x usage back to the standard 5x once May 31 ends. Two of the most-used AI coding tools resetting how they charge inside the same 24 hours is not a coincidence. It's the whole category admitting that a flat monthly fee never matched what an agent actually costs to run.
If you write code with one of these every day, the question isn't whether to panic. It's whether you've ever once looked at how many tokens your own workflow burns. Most solo operators haven't, because flat pricing meant they never had to.
What actually changes on the Copilot side
The seat prices don't move. Copilot Pro stays $10/month, Pro+ stays $39, Business stays $19/user, Enterprise stays $39/user. What changes is what those dollars buy. Instead of counting "premium requests," every plan now includes a monthly allotment of GitHub AI Credits, and once you burn through the allotment, you pay for usage (input, output, and cached tokens) at the listed API rate for whichever model you picked.
The detail that decides whether this hits you: code completions and Next Edit suggestions stay included and don't consume credits. So if your Copilot usage is mostly inline autocomplete, your bill is functionally unchanged. The people who'll feel it are the ones running agent mode all day: multi-file edits, long agentic sessions, big-context refactors. That's exactly the usage GitHub was subsidizing under flat pricing, and exactly the usage they've now metered.
This is rational. It's also a tax on the most productive way to use the tool, which is worth sitting with for a second.
What changes on the Codex side
OpenAI inserted a $100/month Codex tier back in April, between the $20 Plus plan and the $200 Pro plan. As a launch promotion, that $100 tier shipped at 10x the Codex usage of Plus. After May 31, it reverts to the standard 5x. Nothing is being taken from you that you paid for (the promo was always dated), but if you signed up in April and calibrated your habits around 10x headroom, your effective ceiling halves overnight.
The $200 Pro plan keeps its 20x-Plus allowance on an ongoing basis, with the higher five-hour limits that ran at 25x through May 31 also settling back down. So the practical question for anyone on the $100 tier is whether your real usage was ever close to the 5x line, or whether the 10x promo just gave you room you weren't using.
The actual decision, not the headline
Here's where it gets useful instead of just being news. Both changes point at the same move: stop treating a coding-agent subscription as a fixed cost and start treating it as metered infrastructure you can profile.
For Copilot, that means three things this week. Pull your usage report: GitHub has been letting users download April activity translated into AI Credits since mid-May, so you can see your projected bill before the meter starts. If your usage is mostly completions, do nothing; you're already on the included side of the line. If you're an agent-heavy user, look at which model your agent runs against, because the token rate is per-model, and a cheaper model on the same task can swing the overage bill more than your seat price.
For Codex, the move is simpler: if you're on the $100 tier, watch your usage for the first week of June. If you bump the 5x ceiling regularly, the $200 tier's 20x might be cheaper than buying overage piecemeal. If you never come close, you were overpaying for the promo headroom and the reset costs you nothing.
The honest take
I don't think usage-based billing is a rip-off, and I'm not going to pretend it is. It's the correct way to price something whose cost is genuinely variable. A flat $10 that includes unlimited agent runs was never going to survive contact with people running agents in a loop overnight.
But there's a real cost that doesn't show up on the invoice: predictability. Part of why a solo operator can run lean is that the monthly stack is a known number. Usage-based pricing trades that for accuracy, and accuracy has a way of arriving as a surprise the first month you do something unusual (a big migration, a weekend of heavy agent work, a client project that doubles your normal volume).
So the thing to actually do this weekend isn't to switch tools or cancel anything. It's to instrument your own burn before the first metered bill lands. Find the usage report, look at the number, and decide with data instead of finding out on July 1. The builders who get burned by usage-based pricing aren't the ones who use a lot, they're the ones who never looked.
Author
Lukas
@lukcombinator