OpenAI has reset usage limits for all ChatGPT Work and Codex users after complaints that GPT-5.6 Sol was burning through quotas faster than expected, according to a July 29 post on X by Tibo, the product lead for OpenAI Codex. The company said it did not reduce allowances in any subscription plan. Instead, its review found several factors behind the higher consumption: Sol tends to work for longer stretches than earlier models, makes more tool calls, coordinates more complex workflows, and uses more tokens at the same reasoning level. OpenAI also said programmatic tool calling in code mode lets Sol run calls in parallel, increasing responses per turn and raising usage. The effect was not evenly distributed. Median users saw acceptable efficiency, while heavy users handling complex tasks burned through limits much faster than expected. OpenAI said it has finished multiple improvements that should extend usage endurance by about 18% in typical scenarios, and it will restore the previously suspended five-hour limit on the following day. The company also acknowledged that stronger capability does not always arrive with better efficiency.
OpenAI has reset usage limits for all ChatGPT Work and Codex users, Tibo, product lead for OpenAI Codex, said in a July 29 post on X, while also addressing complaints that GPT-5.6 Sol was consuming quotas too quickly.
Why OpenAI says usage was draining faster than expected
Over the past several weeks, many users had reported that Codex limits were being used up faster than expected. OpenAI said it had not reduced the allowance attached to any subscription plan. After investigating, the company identified several causes.
According to OpenAI, Sol is more likely than previous models to work for longer periods, make more tool calls, and coordinate more complex workflows. At the same reasoning level, it also uses more tokens. OpenAI added that programmatic tool calling, or code mode, allows Sol to execute tool calls in parallel, which leads to more responses per turn and higher overall usage.
The company said the impact was highly uneven across users. Median users still saw acceptable efficiency, but heavy users handling complex tasks were burning through their quotas much faster than expected.
Improvements completed, five-hour limit to return
OpenAI said it has already completed multiple improvements and expects them to extend usage endurance by about 18% in typical scenarios. It also said the five-hour limit that had been paused during the investigation will be restored the following day.
OpenAI acknowledged the issue directly, saying, “Capability and efficiency do not always improve in sync. Some problems only become visible after large-scale real-world use, and we should have realized that sooner.”
This article was originally published by Bit.Fan. For more cryptocurrency news and market insights, visit www.bit.fan. Disclaimer:
The market information, project data, and third-party content displayed on this platform are for industry information sharing only and do not constitute any form of investment advice or return commitment.
Cryptocurrency trading carries high risks. Users should fully assess their risk tolerance and make independent decisions. All profits, losses, and legal responsibilities are borne by the users themselves.
