GitHub Copilot: Rising Token Costs Shock Users

Since June 1, 2026, GitHub Copilot has introduced a token-based billing system, changing the cost of using this artificial intelligence service for software development. While monthly subscription rates remain unchanged, users must now manage a limited number of monthly credits, leading to increased costs for many.

A Game-Changing Credit System

Subscriptions to GitHub Copilot, such as Copilot Pro at $10 per month, Business at $19 per user, and Enterprise at $39 per user, now include a fixed number of credits. For example, an Enterprise user receives 3,900 credits per month, while a Business user has 1,900 credits. These credits are consumed based on the silicon effort expended by the AI, and costs vary depending on the model used. For instance, using ChatGPT-5.2 costs $1.75 per million input tokens, while output credits are charged at $14 per million tokens. Cached inputs are priced at $0.175 per million tokens.

Code completions in a developer's integrated development environment (IDE) and "next edit" suggestions remain free, but code review processes are charged at the same rates as other GitHub Copilot activities. Once their allocated credits are exhausted, users have the option to purchase more.

Unhappy Users

User feedback on this new pricing is abundant. Some report that their credits are depleting much faster than expected. User rvs99 noted that 12% of their credits were used for a minor task, while user prhost saw their 7,000 credits cut in half in just one day. User zoomp05 expressed the general sentiment by stating that GitHub's strategy was clear, but it would have been better to announce it as a subsidized trial.

GitHub's initial subscription offerings were likely viewed as loss leaders by Microsoft, but operating a large language model (LLM) entails additional costs related to developing new models, maintenance, building data centers, and repaying future loans.

What Alternatives for Businesses?

In light of these changes, businesses must reassess the return on investment (ROI) that AI coding platforms provide and adjust budget allocations accordingly. They may consider reorganizing their workflows to reduce costs or explore less expensive alternatives. Open-source models hosted on-premises are not cutting-edge LLMs and lack many features of professional coding platforms. Near-state-of-the-art models are hosted by LLM providers such as Huawei and Alibaba. Secondary coding platforms like Cursor may offer temporary relief, although they often utilize cutting-edge models from OpenAI and Anthropic, and could adopt similar billing practices to GitHub Copilot.