OpenAI Resets ChatGPT Work and Codex Usage Limits, Aims to Extend Usage by 18%

ChainCatcher report: On July 29, Tibo, Head of Product for OpenAI Codex, announced on X that usage limits for all ChatGPT Work and Codex users have been reset, and provided an explanation regarding the rapid consumption of GPT-5.6 Sol tokens. For several weeks prior, numerous users reported that Codex usage limits were being depleted faster than expected. OpenAI emphasized that no subscription plan’s token allowance has been reduced. Investigations revealed multiple causes: Sol is more inclined than previous models to engage in prolonged tasks, perform more tool calls, and coordinate complex workflows; it consumes more tokens at the same reasoning level; programmatic tool calls (code mode) enable Sol to execute tool calls in parallel, resulting in more responses and higher usage per round; usage distribution is highly uneven—median users experience reasonable efficiency, but heavy users handling complex tasks deplete their quotas far faster than anticipated. OpenAI has completed several improvements and expects a roughly 18% extension in usage duration under typical use cases. The five-hour limit, which was paused during the investigation, will be reinstated tomorrow. OpenAI acknowledged, "Capabilities and efficiency do not always improve in tandem; some issues only become apparent after large-scale real-world usage, and we should have recognized this sooner."