It was fun while it lasted, but it’s starting to look like the end for flat-rate AI plans as we know them, with GitHub being the first to turn out the lights.
Just a week after announcing it was halting signups for its flat-rate Copilot Pro and Pro+ plans, Github has announced that starting in June, those plans will switch over to usage-based pricing.
Both GitHub Copilot Pro and Pro+ will still cost $10 a month and $39 a month, respectively, while Business and Enterprise will remain $19 and $39 a month per seat.
But beginning June 1, those plans will replace a fixed allotment of “premium requests units,” which are based on a user’s AI request count and adjusted based on the strength of the model, with “AI Credits,” which are based on the actual tokens used during AI exchanges.
Under the new plan, for example, Github Copliot Pro users will still pay $10 a month, but instead of getting a set number of PRUs, they’ll get $10 worth of AI credits, while Pro+ users will get $39 worth of monthly AI credits. A similar AI credit allotment will apply for Business and Enterprise users.
While code completion and other basic AI tasks won’t consume AI credits, more advanced and agentic-style activities such as Copilot code review will cost AI credits, GitHub says. Users who spend all their AI credits before the month is up will have the option to buy more.
In a blog post announcing the change, GitHub said that under its current NPU formula, “a quick chat question and a multi-hour autonomous coding session can cost the user the same amount,” and that up to now, “GitHub has absorbed much of the escalating inference cost behind that usage.”
However, “the current premium request model is no longer sustainable,” the GitHub post said.
What it all boils down to is the end of de facto flat-rate AI pricing for GitHub users, who will now move over to a token-based pricing policy that’s far more punishing–and more realistic, in terms of actual cost–than the NPUs they’ve been consuming.
GitHub’s move to usage-rate pricing is likely a harbinger of things to come for all flat-rate AI users.
The truth is that the flat-rate plans from Anthropic, Google, and OpenAI have long been loss leaders, devised to grow their user bases and get new subscribers hooked on their AI-powered wares.
Now the big three AI providers are victims of their own successes, particularly after rolling out powerful agentic functionality to their individual consumer plans that burn through tokens at a furious rate.
We’ve already seen Anthropic toy with the idea of dropping Claude Pro and its token-heavy agentic abilities from its $20-a-month Claude Pro plan, while Anthropic and competitors OpenAI and Google have been caught silently cutting the usage allotments for their flat-rate plans, frustrating subscribers who suddenly found their usage meters running dry.
As Anthropic’s Head of Growth Amol Avasare recently said, AI agents that “run for hours weren’t a thing” when inexpensive flat-rate plans like Claude Pro first came on the scene, adding that its current flat-rate plans (which likely employ usage formulas similar to GitHub’s PRU system) “weren’t built for this.”
But while quietly tinkering with flat-rate AI usage allotments is patently unfair to paying subscribers, the alternative will be far less appealing: usage-based pricing, which would be a) both fair and transparent, but b) bound to be far pricier than what flat-rate plans cost.
Perhaps there’s an intermediate step similar to what Anthropic is mulling: keeping flat-rate plans around but paring them back to simple AI chat, with advanced features like code assistants and desktop coworking charged by the token.
Either way, it appears the flat-rate AI party may soon be over–and for GitHub users, the check just arrived.
