Back to Blog
GitHub Copilot usage-based billing

GitHub Copilot's New Usage-Based Billing (June 2026): What It Actually Costs You

ByT-Minus AI EditorialJune 11, 20267 min read
GitHub Copilot's New Usage-Based Billing (June 2026): What It Actually Costs You

On June 1, 2026, GitHub Copilot changed how it charges. The old model gave each plan a fixed monthly allotment of "premium requests." The new model gives each plan a pool of AI Credits, and your usage is metered by token consumption — input, output, and cached tokens — at each model's API rate. Same monthly price for most plans, but what you get for it now depends entirely on which models you use and how heavily.

Fast answer

  • Copilot now bills by AI Credits (token usage), not a fixed premium-request count.
  • Copilot Pro is $10/mo and includes $10 in monthly AI Credits. Pro+ is $39/mo with $39 in credits. Business is $19/user/mo; Enterprise is $39/user/mo.
  • Heavy models (top-tier reasoning) drain credits far faster than lightweight ones — the same monthly fee buys very different amounts of work depending on model choice.
  • Once you exhaust your included credits, paid plans can buy more usage; otherwise you drop to the base model until the next cycle.

What actually changed

Under the old system, "premium requests" were a countable unit — you got, say, 300 a month, and you could roughly predict when you would run out. Under usage-based billing, a single complex request to a top-tier model can consume many times the credits of a simple one. GitHub frames this as fairer (you pay for what you use), and developers frame it as harder to predict. Visual Studio Magazine summed up the community reaction bluntly: "you will get less, but pay the same price."

The practical effect: your $10 or $39 no longer maps cleanly to a number of requests. It maps to a dollar pool of credits that drains at different speeds depending on what you ask and which model answers.

The plans, in plain terms

  • Copilot Free: a limited monthly allowance (around 2,000 completions) — fine for occasional use, not for daily work.
  • Copilot Pro ($10/mo): includes $10 of AI Credits. The default individual plan; light-to-moderate daily use.
  • Copilot Pro+ ($39/mo): includes $39 of AI Credits plus priority access to new models. For heavy individual users who lean on top-tier models.
  • Copilot Business ($19/user/mo): team controls, policy management, and a per-seat credit pool.
  • Copilot Enterprise ($39/user/mo): everything in Business plus a larger credit pool and enterprise-grade controls.

How to not get surprised by overages

  • Know which model is answering. Top-tier reasoning models drain credits fastest; defaulting to a lighter model for routine completions stretches your pool dramatically.
  • Watch your burn rate, not your request count. The old "I have X requests left" intuition is dead — what matters now is dollars-of-credits consumed per day.
  • Set a mental (or real) budget. If your plan includes $10 of credits across ~22 working days, that is roughly $0.45/day. Blowing past that early in the month means base-model-only later.
  • Reserve heavy models for hard problems. Use the cheapest capable model for boilerplate and autocomplete.

FAQ

What are GitHub Copilot AI Credits?

AI Credits are a dollar-denominated pool included with each Copilot plan (e.g. $10 of credits on Pro). Your usage is metered by token consumption at each model's API rate and deducted from that pool. When it runs out, paid plans can purchase more or fall back to the base model.

Did GitHub Copilot get more expensive in 2026?

The headline monthly prices are largely unchanged ($10 Pro, $39 Pro+), but because billing is now usage-based, many developers effectively get less work per dollar if they rely on top-tier models — which is why the community reaction was that you "get less but pay the same."

How do I track my Copilot AI Credit usage?

GitHub shows usage in your billing settings, but it is after-the-fact. To watch your burn rate live and forecast when you will run out, a menu bar tracker like Tokens 4 Breakfast reads your usage in real time alongside your other AI tools.

Comparing Copilot against Cursor and Claude Code?

See all three coding agents ranked by workflow fit and cost in the AI Picks hub.

Open the AI Tools Hub

Sources