OpenAI Resets GPT-5.6 Sol Usage Limits, Quota Duration to Be Extended by Approximately 18%

TechFlow reports, July 29, OpenAI has reset usage quotas for all ChatGPT Work and Codex users and updated the usage limit instructions for GPT-5.6 Sol. Officials emphasized that no subscription plan quotas have been reduced. Addressing previous user feedback that Sol was consuming Codex quotas too quickly, OpenAI has completed multiple optimizations. It is expected that under typical usage scenarios, quota duration will be extended by approximately 18%, with some users already experiencing significant improvements starting today; the previously temporarily suspended 5-hour limit will also be restored the next day.

OpenAI stated that Sol tends to engage in longer reasoning, initiate more tool calls, and coordinate cross-tool and sub-agent workflows in complex tasks, therefore potentially resulting in higher token consumption when solving high-difficulty problems. In addition, the programmatic tool call mode enhances capabilities for parallel processing and continuous execution during waiting periods, but also leads to higher-than-expected single-turn response counts, cached input tokens, and overall usage. Related issues are particularly evident when waiting for tool returns or executing large volumes of web searches. Officials stated that optimizations have been made regarding this and will continue to improve code mode efficiency.

OpenAI also pointed out that this impact is not evenly distributed among users: median users actually consider Sol's token efficiency to be relatively high, but heavy users handling high-complexity tasks consume quotas significantly faster. Officials admitted that before release, they focused too much on average and median performance, neglecting long-tail high-consumption scenarios.