When did GitHub Copilot switch to usage-based billing?

On June 1, 2026. GitHub replaced flat premium-request allowances with usage-based AI Credits, so usage beyond a seat's monthly allotment is billed per token. Base seat prices were not changed. The shift is described in GitHub's own announcement.

How is GitHub Copilot billed now: seat, credits, or tokens?

All three layers at once. You pay a base seat (roughly $10 to $39 a month depending on plan), which includes a monthly allotment of AI Credits; usage beyond that allotment is billed per token (input, output, and cached), priced model by model. Plans and allowances change, so treat GitHub's pricing page as the source of truth.

Can I use GitHub Copilot without paying usage charges?

Largely, if you stay within the AI Credits allotment included with your seat. Charges accrue once usage runs past that allowance, and agentic, multi-file, long-context work is what draws credits down fastest. Reserving heavy agentic runs for work that actually ships is one practical way to hold the bill down.

Is GitHub Copilot cheaper than direct Anthropic or OpenAI API access?

It depends on your usage. After the change, some developers said they would move to direct model access or routers like OpenRouter to control token spend. Whether that is actually cheaper turns on how much you use and how much of the output you keep, which is a cost-per-outcome question, not a sticker-price one.

How do I know if GitHub Copilot is worth the new cost?

Stop at the bill and you cannot tell. Worth it is cost per outcome: how much of Copilot's output ships and survives in your repository per dollar. Codelitics measures the survival of Copilot-authored code and a cost per realized change, so the metered spend can be judged on what it kept, not on what it consumed.

GitHub Copilot billing

GitHub Copilot usage-based billing and AI Credits, explained

On June 1, 2026, GitHub Copilot moved from a flat per-seat subscription to usage-based, token-metered billing built around AI Credits. This is a plain-English explainer of what changed, how AI Credits work, why some developers saw bills jump 10x to 50x, and how to decide whether Copilot still earns its cost.

Measure your GitHub Copilot ROI·How to measure AI coding ROI

What changed on June 1, 2026

From a flat seat to a base fee plus a meter

Copilot used to be largely a flat per-seat subscription with fixed allowances. On June 1, 2026, GitHub moved to usage-based billing: each paid seat still has a base price (roughly $10 to $39 a month depending on plan), but it now comes with a monthly allotment of AI Credits, and usage beyond that allotment is billed per token.

GitHub framed the change as aligning price with the compute that modern, agentic Copilot consumes: the more the models work on your behalf, through multi-step agents and long context, the more it costs to run. That is a defensible reason for the model. It also moves the question from "is the seat active" to "what did the metered usage produce."

AI Credits

What GitHub Copilot AI Credits are, in plain terms

AI Credits are the usage currency. Your seat includes a monthly allotment of them; once you pass it, consumption is billed per token (input, output, and cached), priced model by model, so a request to a larger or newer model draws credits down faster than a smaller one. Agentic and long-context work consumes the most.

How a GitHub Copilot bill is built under usage-based billing: base seat, an included monthly AI Credits allotment, and per-token usage overage.
Component	How it works
Base seat	A fixed monthly price per user (roughly $10 to $39 depending on plan).
Included allotment	A monthly allowance of AI Credits bundled with the seat.
Usage overage	Anything beyond the allotment, billed per token (input, output, and cached), priced model by model.

Plans, allowances, and per-model rates change, so treat GitHub's own pricing as the source of truth for the live numbers.

The 10x to 50x shock

Why some GitHub Copilot bills jumped 10x to 50x

Because the meter rewards restraint and charges for heavy agentic use, power users felt it first. After the change, developers reported agentic bills jumping 10x to 50x: one from about $29 to $750 a month, another from $50 to $3,000. The seat price did not move; the metered usage on top of it did.

The reaction was loud. Some developers said they would leave Copilot for direct API access to Anthropic and OpenAI models, or routers like OpenRouter, arguing they could control token spend better themselves. Whether direct access is actually cheaper depends entirely on how much you use, and on how much of that usage produces code you keep. That is a cost-per-outcome question, not a sticker-price one.

After the price change

So is GitHub Copilot still worth it in 2026?

The honest answer is that the sticker price was never the right question, and the meter makes that obvious. What matters is cost per outcome: of everything Copilot writes, how much actually ships, survives in your repository, and was worth the tokens. A high bill on code that all ships is cheap; a low bill on code that all gets reverted is expensive.

That is the measurement the new pricing forces. Instead of asking "is Copilot too expensive," ask "what did the metered spend ship and keep." It is the same gap between perceived and real value that a METR controlled study found when experienced developers were measured about 19% slower while believing they were faster.

Codelitics measures exactly that: the survival of Copilot-authored code and a cost per realized change, across every AI tool you run. See how to measure GitHub Copilot ROI, why a token dashboard is not the yield, and how to govern AI coding spend on yield rather than raw tokens.

GitHub Copilot billing FAQ

Questions developers ask about the new Copilot pricing

When did GitHub Copilot switch to usage-based billing?: On June 1, 2026. GitHub replaced flat premium-request allowances with usage-based AI Credits, so usage beyond a seat's monthly allotment is billed per token. Base seat prices were not changed. The shift is described in GitHub's own announcement.
How is GitHub Copilot billed now: seat, credits, or tokens?: All three layers at once. You pay a base seat (roughly $10 to $39 a month depending on plan), which includes a monthly allotment of AI Credits; usage beyond that allotment is billed per token (input, output, and cached), priced model by model. Plans and allowances change, so treat GitHub's pricing page as the source of truth.
Can I use GitHub Copilot without paying usage charges?: Largely, if you stay within the AI Credits allotment included with your seat. Charges accrue once usage runs past that allowance, and agentic, multi-file, long-context work is what draws credits down fastest. Reserving heavy agentic runs for work that actually ships is one practical way to hold the bill down.
Is GitHub Copilot cheaper than direct Anthropic or OpenAI API access?: It depends on your usage. After the change, some developers said they would move to direct model access or routers like OpenRouter to control token spend. Whether that is actually cheaper turns on how much you use and how much of the output you keep, which is a cost-per-outcome question, not a sticker-price one.
How do I know if GitHub Copilot is worth the new cost?: Stop at the bill and you cannot tell. Worth it is cost per outcome: how much of Copilot's output ships and survives in your repository per dollar. Codelitics measures the survival of Copilot-authored code and a cost per realized change, so the metered spend can be judged on what it kept, not on what it consumed.

Keep reading: measure GitHub Copilot ROI, compare a Copilot analytics alternative, govern AI coding spend, or start from how to measure AI coding ROI.