GitHub Copilot Code Review Shifts to Metered Billing: New Token-Based Pricing Model Raises Cost Predictability Concerns

Summary

GitHub has announced significant changes to its Copilot pricing model effective June 1, 2026. Starting today, Copilot features including Code Review, Chat, CLI, and agents no longer come with subscription-included allocations but instead consume from a shared organization-wide pool of GitHub AI Credits. One AI Credit equals $0.01, with consumption metered by input tokens, output tokens, and cached tokens at published per-model rates.

For existing Copilot Business customers, a promotional rate of 3,000 AI credits per user per month runs through September 1, 2026, then drops to 1,900 credits per user monthly. Critically, Copilot Code Review introduces a dual billing structure: it consumes AI Credits for the review itself but also charges separate GitHub Actions minutes at standard rates for private repositories, creating distinct line items that do not integrate.

The new model eliminates the previous fallback behavior where users could continue with a lower-cost model once budgets were exhausted. Now, when the credit pool runs dry, premium features stop entirely until the entitlement resets or additional credits are purchased. Code Review is particularly vulnerable to cost overruns since token consumption depends on model reasoning time and the specific model remains undisclosed, making advance cost estimation impossible.

The shift reflects industry-wide trends as major AI providers seek faster returns on massive infrastructure investments. However, large enterprises face new operational risks: a few engineers running intensive multi-agent sessions could exhaust the entire organization's monthly allocation, potentially locking down all Copilot features until additional approvals and purchases are completed.

• Token consumption remains unpredictable as the underlying model is undisclosed, making cost forecasting impossible for enterprise customers