Table of Contents
GitHub Copilot Pricing in 2026: Which Plan to Pick Before June 1
Five plans. A billing model that flips on June 1. And a model upgrade that already landed.
On May 17, GitHub quietly swapped GPT-4.1 out as the default base model for all Copilot Business and Enterprise organizations, replacing it with GPT-5.3-Codex (per the GitHub Changelog, May 17). Then on June 1, 2026, the billing system itself changes: premium request units (PRUs) go away, replaced by GitHub AI Credits billed by token consumption. GPT-4.1 deprecates the same day.
If you're staring at the "Pricing update coming soon" banner on github.com/features/copilot/plans and trying to figure out what you're actually paying for, here is the full picture, prices pulled from that page on 2026-05-26.
The five plans at a glance
| Plan | Price (plans page) | Premium requests/mo | Cloud agent | Best for |
|---|---|---|---|---|
| Free | $0 | 50 | No | Occasional use, personal projects |
| Pro | $10/user | 300 | Yes | Daily development, full agent access |
| Pro+ | $39/user | 1,500 | Yes (incl. third-party agents) | Heavy agentic sessions, model flexibility |
| Business | $19/user | Per seat | Yes | Teams needing policy controls + IP indemnity |
| Enterprise | $39/user | Per seat | Yes | Large orgs needing fine-tuning + SAML SSO |
Prices from github.com/features/copilot/plans, fetched 2026-05-26. Note: upgrades to Pro and Pro+ are currently paused while GitHub rolls out its new flexible billing experience.
What changes on June 1 (GitHub AI Credits)
Premium requests are replaced by GitHub AI Credits
The current system counts "premium request units." One agentic session, one unit (roughly). Starting June 1, Copilot will bill by token consumption at published API rates per model. GitHub calls this currency "GitHub AI Credits" (per GitHub blog, April 27, 2026).
The key pricing shift for individuals: Pro subscribers get $10 in monthly AI Credits. Pro+ subscribers get $39. The seat price doesn't change. What changes is that a quick chat and a two-hour agentic session no longer cost the same unit. The agentic session will draw more credits because it consumes more tokens. (Per GitHub's billing transition post.)
Business and Enterprise plans also stay at their current seat prices ($19 and $39 respectively). As a transition incentive, GitHub is giving Business customers $30 in monthly AI Credits for June, July, and August (vs. the standard $19 equivalent). Enterprise gets $70 for the same window. (Per GitHub's billing transition announcement.)
Token-based billing: what it means in practice
A "token" is roughly three-quarters of a word. A short Copilot chat question and answer consumes a few hundred tokens. A multi-step cloud agent session that reads files, writes code, and runs tests can consume tens of thousands. Under the old PRU system, GitHub absorbed the difference. Under AI Credits, you see it in your bill.
Code completions and Next Edit suggestions are exempt. They still don't consume AI Credits or premium requests under either system. The billing change only touches chat, agent mode, code review, and cloud agent sessions.
If you want to see your projected June bill before it hits, GitHub is rolling out a preview bill via your Billing Overview page on github.com. Check it before June 1 so there are no surprises.
The fallback is gone
Under the PRU system, users who hit their monthly limit could fall back to a lower-cost model and keep working. After June 1, that fallback ends. When credits run out, usage stops unless admins have enabled paid overage or you purchase more credits. For Business and Enterprise, admins get budget controls at the enterprise, cost-center, and user level to cap spend.
GitHub Copilot Free: when it is enough
Free gives you 50 agent/chat requests per month and 2,000 code completions, with no credit card required. Access covers a wide set of models including Claude Haiku 4.5, GPT-5 mini, and a full list of premium models that count against the 50-request cap. (Per github.com/features/copilot/plans, fetched 2026-05-26.)
The ceiling is real. Fifty requests goes fast if you're using Copilot Chat for anything beyond occasional questions. A developer doing a focused debugging session on a gnarly bug can burn through 10-15 requests in an afternoon.
Free excludes the cloud agent, code review, and Copilot CLI. If your use is genuinely occasional, a couple of times a week for boilerplate or quick Q&A, Free works. Any regular daily workflow will hit the wall.
Free is also the right call for developers already assigned a Business or Enterprise seat by their org. Those users aren't eligible for Free as an individual plan, but they get the org plan's full allocation instead.
GitHub Copilot Pro ($10/mo): the default choice for most developers
Pro is the workhorse plan. At $10/user/month, it adds unlimited inline suggestions, 300 premium requests per month, the cloud agent, code review, and Copilot CLI. (Per github.com/features/copilot/plans, 2026-05-26.)
What the 300 premium requests actually cover
Three hundred requests is more than most developers burn in a month on focused coding work. Chat replies, code review on a single PR, a handful of cloud agent tasks. A developer who uses Copilot as a coding assistant (not an autonomous agent running for hours) will typically land well under the cap.
The calculus changes for agentic work. A cloud agent session that researches, plans, and drafts code for a medium-complexity issue can cost 5-10 premium requests depending on model and session length. At that rate, 300 requests is roughly 30-60 agentic sessions a month. Most developers won't hit that.
If you do go over, additional premium requests are available at $0.04 each (per the GitHub Copilot plans page footnote, 2026-05-26). After June 1, that overage converts to AI Credits pricing instead.
Cloud agent and code review are the upgrade triggers
These two features don't exist on Free. The cloud agent lets you delegate a GitHub issue to Copilot, which researches, plans, and drafts a pull request on its own. Code review gives you automated pull request feedback on github.com and in-editor diff review. Both are included in Pro.
The Copilot app (in technical preview) adds parallel session support on top of the cloud agent. If you're planning to run multiple concurrent agent sessions, that guide covers setup.
Auto model selection routes tasks to the right model at a 10% discount
As of May 20, 2026, VS Code's "Auto" model setting routes each request to the optimal model based on task complexity and real-time model availability (per GitHub Changelog, May 20). Paid subscribers get a 10% discount on the multiplier when Auto is selected. A 1x-multiplier model draws 0.9 premium requests instead of 1.
That discount compounds across a month of daily use. It's not dramatic, but it stretches the 300-request budget without requiring you to manually pick a cheaper model. (Per GitHub Changelog, May 20.)
The cloud agent also got smarter about model selection on May 18. Claude Haiku 4.5 and GPT-5.4-mini are now available for straightforward tasks at a 0.33x multiplier (per GitHub Changelog, May 18). Simple changes now cost one-third of a premium request.
GitHub Copilot Pro+ ($39/mo): when premium model access matters
Pro+ costs $39/month and gives you 1,500 premium requests (five times Pro's 300), access to third-party coding agents (Claude by Anthropic and OpenAI Codex in cloud agent, currently in preview), and access to GitHub Spark. (Per github.com/features/copilot/plans, 2026-05-26.)
After June 1, Pro+ subscribers get $39 in monthly AI Credits against the same model pricing. (Per GitHub's billing transition post.)
Who actually benefits from Pro+
A senior developer running multi-hour agentic sessions daily. If you're regularly delegating complex, multi-file refactors to the cloud agent, burning through a 1x-multiplier model at high volume, 1,500 requests gives you five times the headroom. At the cloud agent's cheaper 0.33x-model rate for simple tasks, you could run thousands of simple delegations per month.
Third-party agent access matters if your workflow centers on Claude Opus 4.7 or OpenAI Codex for complex tasks. Pro+ is the plan that integrates them directly into the cloud agent pipeline.
When Pro+ is not worth the 4x price
The upgrade is a $29/month jump (Pro at $10 vs. Pro+ at $39, per github.com/features/copilot/plans). If you're not bumping the 300-request cap and don't need third-party agent delegation in the cloud agent, Pro+ adds nothing you'll notice. Model access for chat and inline suggestions is identical between Pro and Pro+.
Run one month on Pro, check your usage, and see if you're hitting 250+ requests. If not, stay on Pro.
GitHub Copilot Business ($19/user/mo): the team decision
Business adds everything individual plans lack for org-level work: policy management, audit logs, IP indemnity, usage controls, and SAML SSO for organizations using GitHub Enterprise Cloud (per github.com/features/copilot/plans, 2026-05-26).
GPT-5.3-Codex as the new default base model
Since May 17, 2026, GPT-5.3-Codex is the base model for all Business organizations (per GitHub Changelog, May 17). This is the model that runs when your org hasn't approved other models through its internal review process. GitHub's data shows GPT-5.3-Codex has a "significantly high code survival rate" among enterprise customers, which means suggested code tends to stay in the codebase rather than getting reverted.
The LTS guarantee is the enterprise-specific value. GPT-5.3-Codex launched February 5, 2026 and is guaranteed available through February 4, 2027. For an enterprise security team that needs a stable, auditable model version, 12 months of guaranteed availability is concrete. Individual plans (Pro, Pro+) don't get this model stability.
Business vs. Pro+: when org features justify the per-seat cost
At $19/user vs. $10/user for Pro, Business costs nearly twice as much per seat (per github.com/features/copilot/plans). The math works in Business's favor only when org-level controls matter.
IP indemnity is the clearest case: Business includes it, Pro does not. For a team shipping code that ends up in customer products, that's not theoretical. Policy management and audit logs let compliance teams trace exactly which model generated which suggestion.
For an engineering manager comparing Copilot Business to Amazon Q Developer, the Copilot Business vs. Amazon Q Developer comparison covers the feature and pricing delta.
June 1 billing note for Business
Business teams get a promotional AI Credits boost for the first three months after the transition: $30/user in June, July, and August vs. the standard $19. After August, it drops back to $19-equivalent per seat. Admins can set spend caps at the enterprise, cost-center, and user level. The pooled credits model means unused credits from one team member's budget can flow to another in the same org. (Per GitHub's billing transition announcement.)
GitHub Copilot Enterprise ($39/user/mo): large org only
Enterprise adds fine-tuning on your codebase, knowledge bases for retrieval-augmented code suggestions, and native GitHub.com chat integration. The extra $20/seat over Business buys those two features. (Per github.com/features/copilot/plans, 2026-05-26.)
Fine-tuning trains a private model on your internal patterns. Knowledge bases index your repositories so Copilot pulls context from your own org's code. Both are irrelevant for startups and mid-market teams. For a large engineering org where developers spend time working around gaps in generic suggestions, they pay off.
Enterprise also gets a more generous AI Credits promo for the transition: $70/user in June, July, and August (vs. the standard $39). (Per GitHub's billing transition announcement.)
Verdict by persona
Solo dev on a budget. Start with Free. If you hit the 50-request ceiling regularly in week one, upgrade to Pro at $10 (per github.com/features/copilot/plans). That's the right call for the vast majority of individual developers.
Senior developer running agentic workflows daily. Pro at $10. Track your monthly request usage for 30 days. If you're consistently at 250+ premium requests, look at Pro+. The 10% auto-selection discount (per GitHub Changelog, May 20) and cloud agent's 0.33x-model routing for simple tasks may solve your budget problem without the $39 tier.
Startup team of 5-20. Business at $19/user (per github.com/features/copilot/plans). IP indemnity alone is worth it once code ships to customers. Policy controls mean you can gate which models junior developers can invoke. The GPT-5.3-Codex LTS guarantee simplifies your security review.
Mid-market engineering org (100+ seats) with internal tooling, custom libraries, or compliance requirements. Enterprise at $39/user (per github.com/features/copilot/plans). Fine-tuning and knowledge bases pay for themselves at scale. If you're debating between CLI-first tools and Copilot for this use case, read the Grok Build vs. Claude Code vs. Cline comparison for context on where Copilot's GitHub-native integration stands relative to terminal-based agents.
Teams evaluating after June 1. The pricing doesn't change. The billing mechanics do. Check your preview bill in Billing Overview before June 1 so you know your actual AI Credit consumption baseline before you plan the annual budget.
Start with GitHub Copilot Pro at $10/month
FAQ
Are upgrades to Pro or Pro+ available right now?
No. GitHub paused self-serve upgrades to Pro and Pro+ while rolling out the new billing experience. The plans page says: "Upgrades are paused as we roll out a flexible billing experience." Existing subscribers are unaffected. The linked blog post at gh.io/indiv-change-blog has the current status.
What happens to unused premium requests after June 1?
Monthly AI Credits don't carry over, same as premium requests today. Annual Pro/Pro+ subscribers stay on the existing premium request plan until expiration, then move to Copilot Free with the option to upgrade to a monthly plan.
Can I set a spending cap on GitHub AI Credits?
Business and Enterprise admins can set budgets at the enterprise, cost-center, and user level. Individual Pro and Pro+ subscribers don't get granular cap controls, but they can track usage via the preview bill in Billing Overview.
Does Copilot Free include the GitHub Copilot app?
No. The Copilot app requires Pro or above (it runs on top of the cloud agent). Copilot CLI is included in Free.
What is the LTS guarantee for GPT-5.3-Codex?
GPT-5.3-Codex is GitHub's first long-term support (LTS) model, in partnership with OpenAI. It launched February 5, 2026 and is guaranteed available through February 4, 2027 for Business and Enterprise users. That's a 12-month window for enterprise security and safety review cycles. Individual plans don't carry an LTS model guarantee.