OpenAI restores 5-hour Codex and Work limits for ChatGPT Plus users
OpenAI’s decision to reinstate 5-hour usage limits for its Codex and Work tools on ChatGPT Plus has triggered criticism from heavy users who feel constrained by time-based caps and unpredictable quota resets. Many argue that the model’s value is high but that slot machine–style timers, sudden changes, and the inability to “burn” a full week’s quota at once make it hard to plan real work, especially for students, developers, and users in lower-income countries. Others point out that API access and rival services (Claude, Cursor, DeepSeek, local and Chinese models) can offer more flexibility or better economics, underscoring how fragile user loyalty is in a rapidly commoditizing AI market.
Scope of the Change
- 5-hour Codex/Work limits per rolling window are back for Plus and Work users.
- Many see this as effectively reintroducing hard caps for serious and corporate users, especially since higher-priced individual tiers aren’t yet generally available on team/business plans.
Impact on Workflow and Usability
- Power users dislike being stopped mid-session and having to wait for the 5-hour window to reset, especially for weekend sprints, exam cramming, or long coding sessions.
- Some note that cache/context expires between windows, so resuming large tasks can consume a big chunk of the next 5 hours.
- A minority say these limits are actually helpful for work–life balance, forcing breaks.
Rate Limiting, Utilization, and “Casino Mechanics”
- Users accept that limits help smooth compute load, but question why usage can’t drain slower during “off-peak” times; others report there effectively are no off-peak hours (limits hit at 2–4am in various US time zones).
- Some argue limits should be an opt-out guardrail for “casual” users rather than mandatory.
- There is extensive comparison to casino or gacha mechanics:
- Timers (5-hour, 7-day) framed as Skinner-box style reinforcement.
- Unpredictable weekly resets and milestone-based early resets are seen as variable rewards that encourage burning usage quickly.
- Others enjoy resets as “free drinks,” but several complain they make planning impossible and create “junkie” behavior patterns.
Pricing, Value, and Alternatives
- Users in poorer countries feel especially squeezed: they’re not casual, but can’t afford higher tiers.
- Multiple comments say subscriptions provide enormous value versus API pricing, sometimes estimating 10–40× cheaper and citing hypothetical API bills in the tens of thousands per month replaced by a few hundred in subscriptions.
- Others still find the price high and suggest cheaper or more generous competitors (Claude, DeepSeek, Ollama/Ollama Cloud, xAI, Chinese providers) or local models as viable options.
- Concern that limits will tighten over time to push users up the pricing ladder, though some would pay for a truly “unlimited” or much higher-cap plan.
API vs Native Apps
- Some ask why heavy users don’t just switch to the API; responses: API is far more expensive and changes what’s economically feasible to automate.
- A few argue the higher API cost is still worth it as a productivity multiplier; others say their usage levels make that unrealistic.