GitHub Copilot pricing change: Is it still worth it?

Starting June 1, 2026, GitHub Copilot is moving to usage-based billing. The subscription prices stay the same, but the economics change for people who rely on chat, agents, CLI, Spaces, Spark, or code review.

This is more than a billing rewrite. Copilot is moving away from a predictable premium-request model and closer to infrastructure-style metering for premium AI features.

Quick answer

If you mainly use code completions, Copilot may still be worth it. If you spend serious time in chat, agents, CLI, Spaces, Spark, or code review, the effective cost can rise much faster than before.

In this post, I cover:

• what changed
• how the old and new billing models differ
• who is most likely to feel the price increase

GitHub Copilot's old pricing model

Before June 2026, Copilot used monthly premium requests for chat and agent-style features. You had a fixed monthly allowance, and each model or feature consumed a published multiplier.

The main paid plans were:

• Individual: Pro for $10/month and Pro+ for $39/month
• Organizations: Business for $19/user/month and Enterprise for $39/user/month

GitHub Copilot's Premium Request model

Premium requests covered features such as Copilot Chat, agent workflows, and code review. The exact cost depended on the model and feature, but the mental model was simple: you spent from a fixed monthly pool.

GitHub Copilot's new billing model

GitHub AI Credits

Premium requests are being replaced by GitHub AI Credits. GitHub defines 1 AI credit as $0.01 USD, and usage is billed from input, output, and cached tokens.

At a high level, that means:

• chat and agent-style features are now metered by token usage
• GitHub splits included usage into Base credits and Flex allotment
• once included credits are gone, you need additional budget to keep using metered features

Credits apply to Copilot Chat, Copilot CLI, Copilot coding agent and cloud agent, Copilot Spaces, Spark, and third-party coding agents connected through GitHub.

Code completions and Next Edit Suggestions are not billed in AI credits. They remain included for all paid plans.

If you run out of included credits and have no extra budget available, metered features do not fall back to a free model. Chat, agents, CLI, Spaces, Spark, and similar experiences stop working until the next cycle or until more credits are available. Code completions and Next Edit Suggestions continue to work.

Copilot code review also has a second cost layer. Starting June 1, 2026, code review running on GitHub-hosted runners will consume GitHub Actions minutes billed at standard Actions rates. Self-hosted runners avoid the Actions-minute charge, but the AI credit cost still applies.

GitHub Individual subscription plan

Plan	Price	Base credits	Flex allotment	Total included usage
Pro	$10/month	1,000	500	1,500
Pro+	$39/month	3,900	3,100	7,000
Max	$100/month	10,000	10,000	20,000

Copilot Free includes 2,000 code completions per month, an AI credit allowance, and auto model selection.

Sources: GitHub Blog, GitHub Docs

GitHub Business and Enterprise subscription plan

Plan	Price	Total included usage	Transition period included usage
Business	$19/user/month	1900	3000
Enterprise	$39/user/month	3900	7000

Business and Enterprise plans have:

• Free unlimited code completions and next edit suggestions for all paid users
• A transition period from June 1 to September 1, 2026 with increased included usage
• A pool of AI credits shared across the organization
• Admin controls to buy additional credits and set user limits

Sources: GitHub Docs

Important gotchas before you decide

• No rollover: unused AI credits expire at the end of the month. That makes the new model feel less like pure usage billing and more like metered access on top of a prepaid monthly allowance.
• No cheap fallback once credits are gone: metered features stop instead of degrading to a free model. Completions still work, but chat and agent-heavy workflows do not.
• Mobile subscribers have a hard limit: if you subscribed through GitHub Mobile on iOS or Android, you cannot purchase additional AI credits.
• Code review is double-metered: it can burn both AI credits and GitHub Actions minutes, which matters a lot more for teams than it does for solo users.
• Annual-plan users need to read the fine print: the extreme multiplier increases apply if you stay on the old request-based billing until renewal. If you switch to usage-based billing earlier, that specific trap goes away.

A few numbers that explain the shift

One reason the change feels harsher for heavy users is simple: frontier-model output is not cheap. GitHub's pricing tables currently list examples such as:

Model	Output price	Why it matters
GPT-5.4	$15 / 1M output tokens	Long verbose chat sessions add up faster than many people expect.
GPT-5.5	$30 / 1M output tokens	Premium frontier output gets expensive very quickly.
Claude Sonnet 4.6	$15 / 1M output tokens	Even mid-to-high tier premium models are no longer “basically unlimited” in practice.
Claude Opus 4.7	$25 / 1M output tokens	Opus-style agent workflows are exactly where overages can become noticeable.

That does not mean every workflow becomes expensive overnight. It does explain why GitHub wants long chats, large contexts, and agent loops priced more like infrastructure than flat-fee software.

Operational caveats worth knowing

If you are using GitHub's billing preview to estimate future costs, treat early exports carefully. GitHub documented:

• omitted 0x model usage in part of April
• duplicate entries later in the month
• missing code review estimates in some reports
• a May fix for briefly overstated costs in some previews

It is also worth updating your tools before drawing conclusions from the UI. GitHub explicitly recommends newer client versions such as VS Code 1.120 and Copilot CLI 1.0.48 so pricing labels and billing terminology match the new system.

Either pay more, or use less: The new reality of AI tool pricing

The broader shift is hard to miss: AI coding tools are starting to look less like flat-rate SaaS and more like metered infrastructure.

That will push developers to be more intentional:

• use smaller models when possible
• ask sharper questions
• avoid wasteful back-and-forth
• keep strong fundamentals instead of outsourcing everything

That may be healthy in the long run. More expensive AI will not magically make people better developers, but it may force a clearer distinction between real time savings and expensive convenience.

Is it time to switch from GitHub Copilot to alternatives?

Maybe, but I would not rush just because the pricing page changed. Other AI tools follow the same basic pattern: deliver value, then charge for it with margin.

You might still get solid value from Copilot under the new model, especially if you keep token usage under control.

For completion-heavy users, Copilot may still be an easy yes. For people who spend hours in chat, agents, or code review, the answer is much less obvious. That is where you should compare Copilot not just to other subscriptions, but to the economics of using APIs or alternative tools more directly.

In the next article, I will look at practical ways to get more value from AI tools without burning through your budget.

FAQ