Codex limits

OpenAI has rolled out significant changes to the rate limits and consumption metrics for its Codex API, the engine powering GitHub Copilot and other code generation tools. The update, which removed a previous 2x usage bonus and further tightened the 5-hour usage window, has resulted in a dramatic and unexpected increase in resource consumption per query. User reports on forums like Reddit indicate that what was previously a minor API call can now consume a substantial portion of the limit, with some developers stating that just two simple prompts now use roughly 75% of their 5-hour allocation. This suggests the effective cost of using the service has skyrocketed, potentially by a factor of 10 to 15 times compared to the old system.

The backlash from the developer community has been swift and severe, with many labeling the new limits as making the service 'useless' for Plus subscribers who rely on it for regular coding assistance. The core complaint is that the practical utility of the API has been crippled; developers cannot complete meaningful work before hitting their caps. This shift fundamentally alters the value calculation for paying customers, who now face a choice between drastically reduced productivity or significantly higher costs. The situation highlights the challenges and user frustration that can arise when AI service providers adjust their commercial and technical parameters post-launch, impacting workflows that have become integrated into daily development practices.