GitHub Copilot: Access Restrictions Spark Outrage

Unexpected Restrictions Disrupt GitHub Copilot Users

Subscribers to GitHub Copilot have recently faced unexpected interruptions with their programming assistant. These outages occurred following the implementation of new limits and the removal of the Opus 4.6 Fast model. GitHub acknowledged a counting bug that artificially inflated the activity of certain accounts, leading to unforeseen blocks. Users have expressed their frustration over these interruptions, which have affected a variety of profiles, from students to professional developers.

Recent Copilot models, which are more resource-intensive, have quickly pushed quotas to their limits. This situation has particularly impacted subscribers on basic plans, where quotas are more restricted, thus exacerbating disparities among users. Interruptions have been reported by students, Pro subscribers, and enterprise developers, who lost access in the middle of test generation or during simple code comments.

More Resource-Intensive Models

The removal of the Opus 4.6 Fast model has forced users to turn to models that require more computational power. Each request sent to these models consumes more resources than previous versions, accelerating quota consumption. Developers have experienced session interruptions after a series of seemingly short requests, such as generating unit tests or refactoring functions.

The actual consumption per request has increased without users being informed, leading to pair programming sessions being cut off mid-way. GitHub has acknowledged the counting issue, but interruptions described in forums indicate a clear rise in consumption related to the models themselves. Developers accustomed to full days of continuous use have seen their quotas dwindle in just a few minutes. Some have attempted to shorten their requests, but recent models use a wider context window, increasing the load even for simple requests.

Students and Pro Subscribers Reach Their Limits Faster Than Others

The tiers applied to different plans create a visible gap between subscribers. In simple terms, the jump from one plan to another is not significant enough. Students have the tightest quota, and Pro subscribers are just above them. The heavier models consume a significant portion of these quotas, leading to rapid cutoffs. Students have lost access after just a few minutes of use while working on short exercises. Pro subscribers have hit their limits in the middle of a refactoring session, without changing their habits.

Only Enterprise subscribers seem to be spared due to wider quota margins. However, frustration is growing among those who are reaching their limit of patience with these recurring interruptions. GitHub has announced adjustments following the bug fix, but this has not been enough to ease tensions among affected users.