What changed in GitHub Copilot’s enterprise pricing this month

The transition to usage-based billing

GitHub Copilot switched to usage-based billing on June 1, 2026. This change replaced premium request units with GitHub AI Credits. One credit equals one cent. Every interaction with a premium model draws from a monthly pool based on token consumption. This consumption includes input, output, and cached tokens.

Subscription prices remain unchanged.

Existing Business and Enterprise customers received promotional credits during the summer transition, where Business customers received $30 per user and Enterprise customers received $70 per user. These promotional amounts dropped back to base subscription rates on September 1. The shift followed an announcement on April 27, 2026. GitHub paused sign-ups for Copilot Pro, Pro+, and Student plans on April 20, 2026. Additionally, self-serve Copilot Business sign-ups paused on April 22, 2026. Users on monthly Pro or Pro+ plans automatically migrate to usage-based billing on June 1. Annual plan users remain on premium request-based pricing until their plan expires. At expiration, they transition to Copilot Free with the option to upgrade to a paid monthly plan. They may also convert to a monthly paid plan before their annual plan expires. Under the new model, users can no longer fall back to lower-cost models after exhausting their limit. Instead, usage depends on available credits or administrator budget controls, and users can also purchase more AI Credits if they need them.

Usage and model economics

Agentic workflows consume many more credits. A single multi-file task can use more tokens than a simple chat question. This occurs because agents read files and iterate through steps. The variable cost of tokens makes budgeting harder for engineering leads.

Code completions and Next Edit suggestions do not consume credits. Chat, agent mode, and code review consume the credit pool. Copilot code review also uses GitHub Actions minutes. Because frontier models charge more per token, an ambitious agent session that relies on a frontier model might consume 1,800 AI Credits in a single afternoon because the model must process massive amounts of input, output, and cached context data tokens daily. A feature touching five components might consume approximately 118 credits.

The cost of a task depends on the model and the token count. Frontier models like Claude Opus 4.8 cost more per token than lightweight models. Prompt size and output length also influence the total bill. For example, a model with a $3 per million input token rate and $15 per million output token rate will consume credits differently than a lightweight model. Code completions remain unlimited on all paid plans.

You should check your usage.

Governance and organizational controls

Credits pool across the organization for Business and Enterprise plans. This eliminates stranded capacity. One team’s usage draws from the same pot as another team.

Admins set budgets at the enterprise, cost-center, and user levels. Organizations also control spending at the organization level. When the pool hits zero, admins decide whether to allow additional usage at published rates or cap spend. New Copilot Business and Copilot Enterprise customers paying by credit card or PayPal will face new requirements starting October 1. These users must pay for each seat before users gain access. All seats added in the middle of a billing cycle incur prorated charges. Revoking a seat does not result in a prorated refund. Copilot Enterprise includes SAML SSO and codebase-aware chat, while Business does not. For organizations using GitHub Enterprise Cloud, adding Copilot costs an additional $21 per user per month. This brings the effective enterprise cost to $60 per user per month.

Deciding between Business and Enterprise tiers now involves weighing the usage cushion provided by different credit pools. An organization of heavy users on Business hits overage territory at roughly half the consumption an Enterprise pool absorbs. This makes Enterprise a potentially more stable option for teams that rely on agentic workflows.

Starting September 28, 2026, GitHub will relaunch Copilot Chat on github.com, Copilot Chat in GitHub Mobile, and Copilot cloud agent as a single, unified Copilot experience. This launch will also change the default review effort level for Copilot code review from "Lite" to "Balanced".

Will administrators find the new granularity useful?