So GitHub just dropped a bomb. Starting June 1, 2026, Copilot is moving to usage-based billing. No more flat rate “use as much as you want” model. Now you get credits, and when they’re gone, they’re gone.
And the developer community? Absolutely furious.
Let me break down what’s actually happening, why it matters, and what I think this means for all of us.
What’s Actually Changing
Here’s the short version. Every Copilot plan now comes with a monthly allotment of GitHub AI Credits instead of the old premium request system. 1 credit = $0.01 USD. Your usage is calculated based on token consumption — input tokens, output tokens, cached tokens — all billed at published API rates per model.
The plan prices themselves haven’t changed:
- Copilot Pro: $10/month → $10 in monthly AI Credits
- Copilot Pro+: $39/month → $39 in monthly AI Credits
- Copilot Business: $19/user/month → $19 in monthly AI Credits
- Copilot Enterprise: $39/user/month → $39 in monthly AI Credits
Looks the same on paper, right? Same price, same number. But here’s the thing — the value you get for that money is wildly different now.
Why Developers Are Losing Their Minds
The backlash has been massive, and honestly? I get it.
Your $39 Can Disappear in One Session
With agentic workflows becoming the norm — long-running, multi-step AI sessions that do actual work across your codebase — a single session can burn through $30-40 in credits. One developer on the GitHub community discussion reported their $39/month subscription ballooned to $400+ in usage. That’s not a typo. Four hundred dollars.
When GitHub says “agentic workflows consume far more resources than the original plan structure was built to support,” what they really mean is: the flat rate was too generous, and heavy users were costing them money.
No Rollover, No Safety Net
Credits reset monthly. Use it or lose it. If you have a slow month, those unused credits vanish. If you have a busy month, you’re hitting walls or paying overages. This feels like those old cell phone plans where unused minutes just evaporated at midnight 😅
Premium Models Got Locked Behind Higher Tiers
Opus models? Gone from Pro plans. You want Claude Opus 4.x? That’s Pro+ territory now. And even on Pro+, the multipliers mean your credits drain significantly faster when using top-tier models.
For reference, here’s what the models actually cost per million tokens:
| Model | Input | Output |
|---|---|---|
| GPT-4.1 | $2.00 | $8.00 |
| Claude Sonnet 4.x | $3.00 | $15.00 |
| Claude Opus 4.x | $5.00 | $25.00 |
| GPT-5.5 | $5.00 | $30.00 |
Use a premium model for anything non-trivial and your $10 or $39 worth of credits won’t last the week.
The “Free” Stuff Is Shrinking
The free model options that made the individual plan worth it? Removed. As one developer put it: “Removing the free models destroys any value in the individual plan.” That’s harsh, but there’s truth in it.
The Silver Lining (If You Squint)
Look, it’s not ALL bad. There are a few things GitHub got right:
- Code completions and Next Edit suggestions remain free across all plans. The basic autocomplete you use hundreds of times a day? Still unlimited. That’s actually huge.
- Pooled credits for organizations — teams can share credits instead of having unused capacity stranded on inactive seats.
- Budget controls — enterprises can set spend caps at org, cost center, and user levels. No surprise $400 bills if admins set it up properly.
- Preview billing — you can see what your usage would cost before the transition hits.
What This Really Means
Here’s my honest take on the bigger picture.
GitHub Is Admitting the Old Model Was Unsustainable
When a handful of requests can cost more than the entire monthly plan price, the math doesn’t work. GitHub subsidized AI coding for two years to build market share. Now that everyone’s hooked, it’s time to charge real prices. Classic strategy.
This Pushes Developers Toward Alternatives
The community discussions are already flooded with people switching to Cursor, Cline with local models, direct API usage, or running DeepSeek locally. When you make the pricing transparent and token-based, developers suddenly realize they can cut out the middleman and go straight to the API for potentially less money.
And that’s the irony. By exposing the token-based pricing, GitHub accidentally showed developers exactly how much margin they’ve been paying for the Copilot wrapper.
The Era of “Unlimited AI” Is Over
This isn’t just a GitHub thing. Every AI company that offered flat-rate pricing is going to face this same reckoning. The compute costs are real. Models are getting more capable but also more expensive. Agentic workflows — the actually useful stuff — consume 10-100x more tokens than simple chat completions.
We’re moving from “all you can eat” to “pay for what you use.” That might actually be more fair in the long run, but the transition is going to be painful.
Individual Developers Get Hit the Hardest
Enterprise teams with pooled budgets and dedicated admins will figure this out. They’ll set up monitoring, cap spend, and optimize model selection. Individual developers and small teams? They’re going to eat surprise bills or hit walls mid-workflow. That’s not a great developer experience.
What Should You Do?
If you’re on Copilot right now, here’s my practical advice:
- Check the preview bill when it drops in early May (should be available now). See what your actual usage would cost under the new system.
- Audit your model usage. Are you using Opus when Sonnet would work fine? That’s a 2-3x cost difference per token.
- Consider alternatives. If you’re a heavy user, direct API access or tools like Cursor might give you more control over costs.
- If you’re on an annual plan, know that you keep premium request pricing until it expires, but you’ll eventually transition to Copilot Free after that. Plan accordingly.
- Set budget limits the moment the new system goes live. Don’t wait for the surprise.
Final Thoughts
GitHub Copilot isn’t dead. But the Copilot we knew — the one where you paid a flat fee and used it without thinking about costs — that’s definitely dead.
What we’re getting instead is a more honest, more transparent, and potentially more expensive tool. Whether that’s a better deal depends entirely on how you use it.
For casual users who mostly rely on code completions? You’ll probably be fine. The basic stuff is still free.
For power users running agentic workflows daily? Start budgeting. Or start exploring alternatives. Because $10 a month for unlimited AI coding assistance was always too good to last.
The real question is: will GitHub’s execution be good enough to justify staying, or will this be the push that fragments the AI coding tool market for good?
I think we’ll know by the end of summer 🤷♂️
What’s your move? Sticking with Copilot or jumping ship? Hit me up on X (Twitter) or LinkedIn. I want to hear how this is hitting your workflow ✌️