The Party's Over: GitHub Copilot Is Charging You for Every Token You Burn

How the world’s most popular AI coding assistant just flipped the script on flat-rate pricing — and what it means for every developer on the planet.

Wait, What Just Happened?

If you’ve been living your best developer life, spinning up autonomous coding agents, running multi-hour AI sessions, and letting GitHub Copilot do the heavy lifting, all for a flat $10 a month, we’ve got some news for you. The free ride is ending.

GitHub announced that starting June 1, 2026, Copilot is ditching its old “premium request unit” (PRU) model and switching to a usage-based billing system powered by something called GitHub AI Credits. In plain English? You now pay for what you actually use. Every token in, every token out, every cached token, it all counts.

This isn’t a small tweak. It’s a fundamental shift in how AI coding tools get priced. And honestly? It’s been a long time coming.

So What Exactly Is Changing?

Let’s break it down without the corporate jargon.

Right now, Copilot subscribers get a monthly bucket of “premium requests.” You dip into that bucket every time you ask Copilot for help. Simple question? One dip. Multi-hour autonomous coding session? Also one dip. That’s the problem, and GitHub finally admitted it out loud.

As GitHub’s Chief Product Officer Mario Rodriguez put it: “Today, a short chat question can cost the user just as much as an autonomous coding session lasting several hours.” That’s not a pricing model. That’s a buffet where everyone pays the same whether they grab a salad or eat the whole kitchen.

Under the new system, here’s how the plans shake out:

Copilot Pro — Still $10/month, but now includes $10 in AI Credits
Copilot Pro+ — Still $39/month, now includes $39 in AI Credits
Copilot Business — Still $19/user/month, with matching credits
Copilot Enterprise — Still $39/user/month, with matching credits

The base prices haven’t changed. But the way those prices translate into usage? Completely different. And if you burn through your credits before the month ends, you can buy more, at the API rate for whatever model you’re using.

One silver lining: code completions and Next Edit suggestions remain free. They won’t eat into your credits at all. So if you’re just using Copilot for basic autocomplete, you’re probably fine. But the moment you fire up an agentic workflow or ask Copilot to review your code? The meter starts running.

Why Is GitHub Doing This Now?

Here’s where it gets interesting. GitHub didn’t just wake up one morning and decide to shake things up. The writing has been on the wall for months.

According to Ars Technica, leaked internal documents reportedly showed that week-over-week costs for GitHub Copilot had nearly doubled since January 2026. That’s not a slow creep, that’s a rocket ship pointed at GitHub’s infrastructure budget.

The culprit? Agentic AI. Tools that run autonomously, spin up subagents, iterate across entire codebases, and sometimes run for hours without stopping. These workflows are incredible for productivity. They’re also absolutely brutal on compute costs.

GitHub’s own blog post on changes to individual plans was refreshingly honest about it: “It’s now common for a handful of requests to incur costs that exceed the plan price.” Read that again. A few requests. More expensive than the entire monthly subscription. That’s not sustainable for anyone.

So GitHub did what any rational business would do. First, they paused new signups for Pro, Pro+, and Student plans. Then they tightened usage limits. Then they pulled Opus models from lower-tier plans. And now, they’re overhauling the entire pricing structure.

The Token Economy: What You’re Actually Paying For

Here’s where things get a little technical, but stick with us, because this part matters.

Under the new model, your AI Credits get consumed based on token usage. Tokens are the basic units of text that AI models process. Every word you type, every line of code Copilot generates, every bit of context it holds in memory, all of it translates into tokens.

The tricky part? Different models cost wildly different amounts. OpenAI’s GPT-5.4 Mini runs at $4.50 per million output tokens. GPT-5.5 costs $30 per million output tokens. That’s a 6x difference. If you’re casually using a lightweight model for quick questions, your credits stretch far. If you’re unleashing a frontier model on a complex refactor? They evaporate fast.

This is why GitHub is also launching a “preview bill” tool in early May, before the new pricing kicks in on June 1. It lets you see exactly how your current usage would translate into credits under the new system. Think of it as a financial reality check before the bill actually arrives.

As Blockchain.news reported, business users get a bit of a cushion during the transition. Organizations on Business plans receive $30/month in promotional credits from June through August. Enterprise users get $70/month in promotional credits during the same period. GitHub is essentially giving companies a grace period to adjust their workflows before the full weight of usage-based pricing hits.

What This Means for Developers Day-to-Day

Okay, real talk. How does this actually affect you?

If you’re a casual Copilot user, someone who uses it for code completions, occasional chat questions, and light assistance, you’re probably going to be fine. Your $10 in monthly credits will likely cover your usage without breaking a sweat.

But if you’re a power user? A developer who runs agentic workflows, uses Copilot to autonomously tackle complex multi-file projects, or relies on the most powerful frontier models for everything? You’re going to feel this change. Hard.

GitHub’s own blog post offered some practical tips for managing your usage under the new system. Use smaller models for simpler tasks. Enable plan mode in VS Code or Copilot CLI to improve task efficiency. Reduce parallel workflows, tools like /fleet burn through tokens fast. And if you’re on Pro and constantly hitting limits, consider upgrading to Pro+, which offers more than 5x the limits.

The good news for organizations is that the new system introduces pooled credits across teams. Instead of individual users sitting on unused quotas while others run dry, credits flow where they’re needed. Admins also get new budget controls, spending caps at the enterprise, cost center, or individual user level. That’s actually a meaningful improvement for finance teams trying to manage AI spend.

Is This the End of Cheap AI Coding?

Let’s zoom out for a second, because GitHub isn’t doing this in a vacuum.

PCWorld put it bluntly: “It was fun while it lasted, but it’s starting to look like the end for flat-rate AI plans as we know them.” And they’re not wrong.

The flat-rate AI subscription model was always a bit of a fantasy. Anthropic, OpenAI, and Google have been quietly trimming usage allotments on their flat-rate plans for months. Anthropic briefly tested removing Claude Code from its $20/month Pro plan. Anthropic’s Head of Growth Amol Avasare even admitted that AI agents that “run for hours weren’t a thing” when those inexpensive flat-rate plans first launched, and that the current plans “weren’t built for this.”

The economics are simple and brutal. AI inference is expensive. Agentic workflows are very expensive. And the more powerful the model, the more expensive every single interaction becomes. Flat-rate plans were loss leaders, designed to hook users and grow subscriber bases. Now those same users are running autonomous agents that cost more per session than the entire monthly subscription.

Something had to give. GitHub just happened to be the first major player to pull the trigger on a full pricing overhaul.

The question now is: who’s next? Anthropic? OpenAI? Google? All signs point to yes. The era of “pay $20 a month and use AI as much as you want” is quietly drawing to a close. Usage-based pricing is the future, and GitHub just gave everyone a preview of what that future looks like.

The Bigger Picture: AI’s Compute Crunch

There’s a broader story here that goes beyond GitHub’s balance sheet.

The AI industry is facing a genuine compute crunch. Demand for AI inference is skyrocketing. The chips to power that inference are in short supply. And the costs of running frontier models at scale are staggering. Every major AI company is grappling with the same fundamental tension: users want unlimited access, but unlimited access is financially unsustainable.

Usage-based pricing is the industry’s answer to that tension. It’s transparent. It’s fair. It aligns what users pay with what they actually consume. But it also means that the most powerful AI tools, the ones that can genuinely transform how developers work, are going to cost more than a Netflix subscription.

For individual developers, that’s a real consideration. For enterprises, it’s a budget line item that’s about to get a lot more scrutiny.

What Should You Do Right Now?

Here’s your action plan before June 1 hits:

1. Use the preview bill tool. GitHub is launching it in early May. Use it. Understand your projected costs before the new system goes live.

2. Audit your workflows. Are you running parallel agentic sessions? Using frontier models for tasks that a smaller model could handle? Now is the time to optimize.

3. Check your plan. If you’re on Pro and regularly hitting limits, Pro+ might actually be worth the upgrade, especially with the 5x higher usage ceiling.

4. If you’re on an annual plan, you’re not immediately affected. Individual developers on annual plans stay on the current PRU-based pricing until their subscription expires. But start planning now.

5. Watch the broader market. GitHub is the first domino. Other AI providers will follow. Understanding how token-based pricing works now will save you a lot of confusion later.

The Bottom Line

GitHub Copilot’s move to usage-based billing isn’t a betrayal. It’s a reality check. The flat-rate AI party was always going to end, the only question was when and who would go first.

The new system is more transparent, more fair, and more sustainable. It rewards light users and asks heavy users to pay their actual share. That’s not unreasonable. But it does mean that the era of unlimited AI coding for $10 a month is officially over.

The check has arrived. Time to see what you actually ordered.