On April 27, 2026, GitHub announced that all Copilot plans are moving to usage-based billing on June 1, 2026. The flat "premium request" model is being retired. Every Copilot interaction will now be billed by token consumption, the same way every major LLM API has always priced. For solo developers using Copilot for autocomplete, this barely changes anything. For engineering teams running agent workflows, AI tooling spend is about to become a real line item that needs budgets, monitoring, and policies.

Key Takeaways

  • Copilot's Premium Request Units (PRUs) are being replaced by GitHub AI Credits (1 credit = $0.01 USD) on June 1, 2026.
  • Every chat and agent interaction will be billed by input, output, and cached token consumption — code completions remain free.
  • Plan base prices are unchanged (Pro $10, Pro+ $39, Business $19/seat, Enterprise $39/seat) and each plan now includes monthly credits equal to its price.
  • Annual subscribers face degraded value — model multipliers go up on June 1 for them only — so it's worth running the prorated math before deciding whether to switch to monthly.
  • Heavy agent users will pay more; light users may pay less. Teams need budget caps, pooled credits, and model-selection discipline to avoid surprise bills.

What is actually changing?

Today, Copilot uses Premium Request Units. You get a monthly bucket of requests, and every chat or agent prompt counts as one — regardless of whether it's a one-line question or a 30-minute autonomous coding session. Starting June 1, 2026, that uniform pricing disappears.

  • PRUs are replaced by GitHub AI Credits at 1 credit = $0.01 USD.
  • Every interaction is billed by token consumption — input tokens, output tokens, and cached tokens.
  • Each model has its own per-token rate, matching published API rates.
  • Code completions and Next Edit suggestions remain free across all plans.
  • Plan base prices are unchanged — Pro $10/mo, Pro+ $39/mo, Business $19/seat, Enterprise $39/seat.
  • Each plan includes monthly AI Credits equal to its subscription price (Pro = $10 in credits, etc.).

The big shift: a quick "explain this regex" question and a multi-hour agentic refactor session no longer cost the same. They never really did at GitHub's end — the money was just being averaged across the user base.

Why is GitHub doing this now?

Read between the lines of the official announcement and the picture is clear. Agentic coding broke the math.

A single autonomous coding session — running across an entire repository, calling tools, iterating, retrying, spawning subagents — can consume thousands of times more compute than a chat question. Under PRUs, both counted as "one request." A handful of power users were costing GitHub more than the plan price covers.

GitHub admitted this directly in their April 21 update on individual plans:

"It's now common for a handful of requests to incur costs that exceed the plan price."

That is not a sustainable business. Token-based billing aligns price with cost, which is the same reason every API provider — OpenAI, Anthropic, Google — has always priced this way. Flat-rate worked when "AI assistance" meant autocomplete suggestions on a few hundred tokens. It does not work when "AI assistance" means a multi-step agent burning through 200,000-token contexts on every run.

There is also a quieter reason: GitHub is preparing the ground for the next generation of agentic features. Cloud agents, async workflows running on GitHub Actions infrastructure, parallelised subagent fleets — these are products GitHub cannot ship at scale on a flat-rate model. Token billing is the foundation that lets them keep adding capability without the unit economics collapsing.

Who wins and who loses?

Wins: Developers who mostly use code completions and light chat. You will likely see no real change. Your $10 Pro plan probably never came close to using $10 in tokens. In fact, GitHub's framing suggests most casual users are already subsidising the heavy users — and that subsidy is what's ending.

Loses: Heavy agent users. If you are running long-trajectory agentic workflows — multi-step refactors, autonomous bug-fixing across files, parallel subagents, full repo reviews — you will feel this. Ironically, the same workflows that gave you outsized value under PRUs are exactly what's getting expensive. The product hasn't changed; the price tag has.

Maybe: Teams. It depends entirely on how your developers actually use Copilot. Some engineering organisations will save money under usage-based billing. Others will hit budget caps mid-month for the first time in their lives, and have to have an awkward conversation with finance about why "AI tooling" became the third-largest line item on the engineering budget.

The annual plan trap

If you are on an annual Pro or Pro+ plan, GitHub is doing something clever and slightly hostile.

You keep your existing PRU-based pricing until your annual plan expires — but model multipliers go up on June 1 for annual subscribers only. Translation: your existing annual plan gets quietly worse on June 1. You will burn through your premium request allotment faster, hitting limits sooner.

You have two paths:

  1. Ride out the annual plan at degraded value, then transition to monthly when it expires.
  2. Convert to a monthly plan early and get prorated credit for the unused portion of your annual.

Neither is obviously better. It depends on your usage pattern. But if you are an annual subscriber, you should sit down before June 1 and run the numbers. Don't ignore this — the math actually matters.

What should you do this month?

For individual developers:

  1. Check your current usage. GitHub is launching a preview bill experience in early May. Look at your Billing Overview on github.com and see what your actual token consumption would have cost you over the past few months. This number alone will tell you whether the change is good or bad news for you.
  2. Audit which models you use. Different models have wildly different per-token rates — sometimes by 10x or more. If you are defaulting to the most expensive premium model for tasks GPT-5 mini could handle perfectly well, that habit is about to cost real money.
  3. Decide on annual vs monthly. If you are on annual, run the prorated credit math before June 1.

For engineering teams:

  1. Set up budgets. Admin-level budget controls let you cap spend at the enterprise, cost center, and user level. Use them. Without caps, one developer running parallel agent workflows can blow through your monthly credits in a week and start hitting your purchase limits.
  2. Pool credits across the org. GitHub now allows pooled included usage instead of per-seat isolation. This is genuinely useful — your light users effectively subsidise your heavy users without anyone changing behaviour. For teams of 10+, this alone is reason to revisit how Copilot is rolled out internally.
  3. Educate your team on model selection. "Use the cheapest model that works for the task" is now a real engineering practice with a real budget impact. This is a culture shift, not just a technical one.
  4. Watch out for Copilot code review. It now consumes both AI Credits and GitHub Actions minutes. Two meters running at once. If you have automated code review on every PR, you are effectively doubling your billing surface for that feature.
  5. Track AI tooling cost as a real line item. If you have been treating AI coding as a free productivity boost, that mental model needs to change. We would recommend giving AI tooling its own budget category, with monthly review — same as you would track cloud spend or SaaS licences.

The bigger picture

This isn't just a Copilot story. Every AI coding tool is heading the same direction — Cursor, Cody, Continue, Claude Code, the whole space. Token-based pricing is becoming the default because flat-rate doesn't survive contact with agentic workloads.

The era of "$20 a month gets you unlimited AI coding" is over. What's replacing it is more honest, but also more demanding: you need to know what your AI usage actually costs, and you need to manage it like any other infrastructure spend.

For solo developers, this mostly means slightly more attention to which model you use and when. Manageable.

For engineering teams, this is a real shift. Treating AI coding as a cost centre with its own budget, monitoring, and policies — instead of an unlimited buffet — is the new normal. The teams that adapt quickly will get the productivity benefits with predictable spend. The teams that don't will get surprise bills, mid-month limits, and frustrated developers.

The good news: the tools to manage this are getting better. Budget controls, usage dashboards, pooled credits, and per-model multiplier visibility all help. The bad news: you actually have to use them. Setting up Copilot used to be "buy seats, hand them out." Now it's "buy seats, hand them out, set budgets, train your team on model selection, monitor usage, adjust monthly."

That's a lot more like managing AWS than managing a Slack subscription. And that is the new reality for AI tooling across the board.

What this means if you are building with AI

A quick note from us: at Lenka Studio, we build AI-powered features and automations for our clients. The shift to usage-based billing on tools like Copilot is a useful preview of what your own AI features will look like at scale.

If you are shipping a product with AI in it — chatbots, content generation, agentic workflows, anything that calls an LLM — you are already on the same per-token economics that GitHub just adopted. The same questions GitHub is now grappling with publicly (which model for which task, how to set user budgets, how to prevent runaway costs from a single power user) are the questions you will need to answer for your own product.

The teams that build AI products well treat token economics as a first-class engineering concern from day one. The teams that bolt AI on without thinking about cost end up with margin problems six months in.

Frequently Asked Questions

When does GitHub Copilot's usage-based billing start?

The change takes effect on June 1, 2026. From that date, every Copilot chat and agent interaction is billed by token consumption rather than as a flat "premium request." Code completions and Next Edit suggestions remain free across all plans.

Will my Copilot subscription cost more under the new pricing?

It depends on how you use it. If you mostly use code completions and light chat, your costs likely won't change. If you run long-trajectory agent workflows — multi-step refactors, autonomous bug-fixing, parallel subagents — your effective cost will go up because those sessions consume thousands of times more tokens than a quick chat question.

What happens to my annual Copilot plan on June 1?

You keep PRU-based pricing until your annual plan expires, but model multipliers increase on June 1 for annual subscribers only. That means your existing plan gets quietly worse and you will hit limits sooner. You can either ride it out or convert to monthly early and get prorated credit for the unused portion.

How can engineering teams control Copilot spend?

Use the new admin-level budget controls to cap spend at the enterprise, cost-centre, and user level. Pool credits across the organisation rather than isolating them per seat. Educate developers on model selection — using the cheapest model that handles the task is now a real budget lever. Treat AI tooling spend as its own line item with monthly review, the same way you track cloud spend.

Is this a one-off Copilot decision or a wider industry trend?

It's an industry trend. Cursor, Cody, Continue, Claude Code, and the rest of the AI coding space are all moving toward token-based pricing for the same reason — flat-rate plans don't survive contact with agentic workloads. The era of unlimited AI coding for a fixed monthly fee is effectively over.

Planning AI features for your product?

If you are mapping out AI features for your product and want to think through the unit economics before you ship, we can help you model the cost up front rather than learn it from the invoice. Get in touch to walk through your roadmap, or take 5 minutes to run a free brand health score to see where AI-driven automation could move the needle for your business.