GitHub Copilot shifts to per-token AI credits; Pro gets 1,000

GitHub will change Copilot billing from a fixed monthly query allowance to per-token AI Credits starting next month. Copilot Pro subscribers at the $10-per-month tier will receive 1,000 credits; GitHub currently values one credit at $0.01. Credits will be consumed by the tokens used in prompts, the model’s outputs and any cached context the model keeps for a session.

A token is roughly three-quarters of a word. That means 10,000 words of text would equal about 12,000 to 13,000 tokens. The company applied the same measure to code: a code base with 10,000 ‘words’—including expressions, statements, variable and function names-would consume about 12,000 to 13,000 tokens if processed in a single query. Both input text and model outputs count toward the token total, and any session context the model retains also factors into consumption.

GitHub noted the number of tokens a single credit buys will vary by the model used, the mix of input and output tokens, the size of the cached context and additional features invoked. Queries routed to frontier, higher-capability models will cost more per token than queries to smaller models.

Inline code completions that appear as you type and Next Edit suggestions will not use AI Credits and will remain free for users.

The credit system replaces the previous fixed-query model while keeping current subscription price tiers. GitHub described the credit approach as a way to align charges with actual compute and context consumption and to give users more control over usage.

GitHub advised that complex queries that analyze long code bases or run multi-agent interactions will draw credits more quickly and may require buying extra credits or upgrading tiers. The company has provided conversion details to subscribers ahead of the change.

GitHub wrote: “The credit-based approach is intended to give more granular control over usage and to align charges with actual compute and context consumption.”