GitHub is tightening the screws on GitHub Copilot’s individual plans: new Pro, Pro+, and Student sign-ups are paused, usage limits are getting stricter, and the most capable Opus model is being pulled from Pro. The company says the changes are meant to protect existing customers from a service that is being pushed harder by agentic coding workflows than the original pricing and capacity model can comfortably handle.
The subtext is simple: AI assistants are no longer just autocomplete with a nicer UI. Long-running, parallel requests are expensive, and GitHub is now openly admitting that a few heavy sessions can chew through resources in ways that threaten reliability for everyone else. That is not a Copilot-only problem; it is the same bill that is coming due across the AI coding market as vendors try to keep ”unlimited” behavior from colliding with real infrastructure costs.
What is changing in GitHub Copilot Individual plans
For now, GitHub says new sign-ups for Pro, Pro+, and Student plans are paused. Existing subscribers stay on board, but the company is making usage guardrails more visible and more restrictive. Pro+ still offers more than 5X the limits of Pro, and users who need more headroom can move up from Pro to Pro+.
There is also a model shake-up. Opus models are no longer available in Pro plans, while Opus 4.7 remains available in Pro+. GitHub also said Opus 4.5 and Opus 4.6 will be removed from Pro+, which is a blunt reminder that model menus are not a permanent feature, just a billing lever with a nicer coat of paint.
How GitHub Copilot usage limits work
GitHub says Copilot now has two limits: session limits and weekly, or 7-day, limits. Both depend on token consumption and the model’s multiplier, which means a heavier model can burn through allowance much faster than a lighter one, even if the request count looks modest.
- Session limits are designed to stop the service from being overloaded during peak usage.
- Weekly limits cap how many tokens a user can consume over 7 days.
- If you hit a weekly limit, you can keep going with Auto model selection if you still have premium requests left.
- Model choice returns when the weekly period resets.
Those screenshots in VS Code and Copilot CLI are the most practical part of the update. They warn users when they are approaching 75% of their weekly usage limit, which is a lot more useful than discovering a hard stop mid-task and pretending the product is at fault for physics.
Refunds, upgrades, and the new guardrails
GitHub is also offering a way out. If the new limits do not work for you, Pro and Pro+ subscribers can cancel and get a refund for the remaining time on their current subscription by visiting Billing settings before May 20. For people already bumping into the ceiling, the company’s preferred answer is clear: upgrade to Pro+ or use smaller-multiplier models and plan mode more carefully.
The move is unsurprising, even if the timing is annoying for users. OpenAI, Anthropic, and other model providers have spent the past year tightening access in different ways as agentic workloads became more common, because the economics of extended multi-step coding sessions are very different from ordinary chat. GitHub is choosing transparency and throttling over pretending every plan can absorb the same workload forever.
The bigger question is whether this becomes the new normal for AI developer tools: fewer vague promises, more metered reality. If Copilot’s usage dashboard works as intended, rivals will copy the idea quickly. If not, expect more frustrated developers, more plan changes, and more polite corporate language for the same old thing: the robots got expensive.