From Flat Subscriptions to Token-Based Billing
GitHub Copilot’s new token-based billing is a usage-based pricing model where users buy access through GitHub AI Credits, which are spent according to the tokens consumed by their prompts and the models they choose. Instead of paying one flat fee for unlimited premium requests, Copilot now ties cost directly to how much AI power you use. This shift replaces the older premium request system that hid much of the underlying compute cost, especially for long-running agentic sessions. Plan prices stay the same, but every plan now includes a monthly bundle of AI Credits. For many users, this means more predictable visibility into how their activity translates into spend, but it also exposes how heavy workloads can exhaust credits fast. Some developers have already reported that their credits “burned like anything” on what they thought were minor tasks, highlighting the learning curve around this change.
Understanding the AI Credits System and Flex Allotments
Under the new GitHub Copilot pricing, each paid plan now includes a mix of base AI Credits and, for individuals, a flex allotment designed for variable usage. Credits are spent on token usage across input, output, and cached data, at per-model rates that GitHub publishes. For individual subscribers, Pro users paying USD 10 (approx. RM46) receive USD 15 (approx. RM69) in monthly credits, while Pro+ users paying USD 39 (approx. RM179) receive USD 70 (approx. RM321). Joe Binder from Microsoft explains that the base credits are permanent, whereas the flex allotment can change “as the economics of AI evolve, including model pricing, new models, and improvements in efficiency.” Code completions and next edit suggestions remain outside this meter and do not consume credits, which protects lightweight autocomplete work. However, heavy chat or long agentic sessions will draw down credits quickly, so users need to watch their dashboards and adjust their habits accordingly.
Copilot Max and Cost Implications for Power Users
To support developers who push Copilot hard, GitHub has introduced Copilot Max, a top-tier individual plan aimed at sustained, high-volume work. Priced at USD 100 (approx. RM460) per month, Copilot Max includes USD 200 (approx. RM920) in monthly AI Credits, split between USD 100 (approx. RM460) in base credits and a USD 100 (approx. RM460) flex allotment. This gives power users more headroom before hitting limits, compared with Pro and Pro+ tiers. Existing Student, Pro, and Pro+ subscribers can upgrade, although new sign-ups remain temporarily paused. In practice, Max is meant for long agentic sessions that span multiple files and repositories, where traditional subscriptions would have hidden escalating inference costs. The catch is that heavy experimentation can still drain credits faster than expected, as some users have already found when seeing thousands of credits disappear in a single day’s work.
Budget Controls for Business and Enterprise Teams
For business and enterprise customers, GitHub Copilot pricing keeps per-seat fees steady at USD 19 (approx. RM87) and USD 39 (approx. RM179) per user per month, respectively, but shifts to shared AI Credits with detailed budget controls. Credits are pooled at the organization level, allowing power users to consume more while lighter users offset overall demand. Unlike individual plans, business and enterprise seats only receive base credits equal to the per-seat price, with no flex allotment, though GitHub is offering temporary promotional credits through August. A new budget system combines universal user budgets, per-user overrides, cost center budgets, and an enterprise-wide budget that caps metered overage once pooled credits are gone. An example from GitHub describes 400 Copilot Business seats at USD 19 (approx. RM87) each, creating USD 7,600 (approx. RM34,960) in fixed license fees and allowing an enterprise metered budget of USD 5,000 (approx. RM23,000), for a maximum bill of USD 12,600 (approx. RM57,960).
Who Wins and Who Pays More Under Usage-Based Pricing?
The move to token-based billing and AI Credits is a clear step away from what many users now see as an implicitly subsidized trial phase. Subscription-style Copilot access often let developers consume far more tokens than their monthly fee could cover, which was never sustainable as models became larger and agentic workloads more demanding. With usage-based pricing, light and moderate users may find their overall costs unchanged or even easier to predict, especially since plan prices remain constant and include bundled credits. Heavy users, however, are discovering that their credits can vanish faster than expected, leading to complaints that projects might not be worth continuing under new constraints. For organizations, the new budget controls can align spend with real value, but they also create complexity: if user-level budgets collectively exceed the shared pool and enterprise budget, some users will be cut off earlier than their individual limits suggest.
