Codex quota resets are turning free usage into a headache

Weekly quota resets for coding agents are supposed to feel like a bonus. For some heavy users, they are starting to feel like a disruption instead.

Subscription coding tools such as Claude Code and Codex typically enforce both 5-hour and weekly usage caps. As the source article notes, those limits serve two obvious purposes: smoothing demand so servers do not get overloaded, and stopping users from concentrating a full month of usage into a single day before canceling.

Providers have adjusted those limits before, including temporary increases and, at times, removing the 5-hour cap. But the current complaint is not about stricter limits. It is about increasingly frequent, poorly signposted resets of the weekly quota.

For users on a $100/month Codex plan, the author argues, one extra weekly reset is effectively worth $25 if they normally consume the full allowance. With expensive frontier models such as Fable 5 and GPT-5.6 Sol, burning through that quota has become easier. The problem is that resets often arrive without warning and are rarely announced through official company channels.

Recommended reading

Hinge founder raises $18M for AI dating startup Overtone

According to the post, OpenAI reset the Codex weekly quota six times in the past two weeks: July 9, July 10, July 10 again, July 14, July 15, and July 17. The author also points to a lesser-known banked reset system for Codex, with resets issued on July 12 and July 13 that can be manually redeemed later but expire within 30 days.

The source says this has become more noticeable after the arrival of Fable 5 and GPT-5.6. Rather than creating goodwill, the unpredictability can leave users feeling they have “wasted” part of the quota they already paid for if a reset lands while they are only halfway through the week’s allocation.

The author describes closely tracking usage, setting phone reminders for exact reset times, and even wishing for an official endpoint to check Codex usage programmatically. That level of monitoring became more frustrating when resets started happening while their remaining weekly quota was still above 50%. In some cases, the weekly reset time also disappears until a new prompt is submitted.

The broader theory is that this is not accidental. The post suggests the burst of resets may be tied to a brutally competitive July for frontier models, with Fable 5, GPT-5.6 Sol, Grok 4.5, Muse Spark 1.1, and Kimi K3 all pushing users to test alternatives across the price-performance curve.

A more cynical interpretation, the author writes, is that frequent resets keep power users busy inside one platform instead of nudging them to try competitors once their normal quota runs out. But if resets become too common, the logic of higher-priced plans starts to break down. In the author’s view, that could push users to downgrade from $100/month to $20/month simply to avoid paying for quota that keeps getting overwritten.