Whether the average user ends up paying more depends largely on the individual user and, to some extent, on who you ask. In the comments section of the GitHub community discussion page that announced the change in April 2026, there are many reports of users finding their credits are being used up much faster than expected. User “rvs99” says, “12% of the total AI credits were consumed on very trivial tasks. I used Claude Sonnet 4.6 as usual, and accordingly I barely updated 2-3 lines out of 6 total files. It cost about $0.35 per line update.” “prhost” posted a screen capture of its account dashboard showing that it had 3,705 of 7,000 credits left after one day of use, saying, “It would be easier to shut down the project.[Microsoft]shot itself in the foot.”
User “zoomp05” summed up the tone of most commentators this way: “The strategy is clear, but to promote the tool they should have said something like ‘this is a subsidized trial’ from the beginning.”
The initial subscription service from GitHub (now defunct) was probably considered a loss leader by the platform’s owner, Microsoft. It quickly became clear that allowing users to write tokens that far exceeded their subscription value was never sustainable. A quick read on the internet apart from announcements and posts from major model providers shows that subscription-based billing as a business model is only temporary. Perhaps surprisingly, many users are surprised that coding platforms now charge at a level that is commensurate with supplier costs. Running an LLM is not a cheap undertaking, especially when you consider the additional costs of developing new models, post-training maintenance, building data centers, and future loan repayments.
