Tracking OpenAI Codex and ChatGPT Work Usage Resets: An Analysis of the Codex Resets Tracker

Key Takeaways

• Rapid User Growth: OpenAI's Codex and ChatGPT Work platforms have reached a significant milestone of 9 million active users, growing quickly from previous 7M and 8M benchmarks.
• Reset Frequency: Usage limits are reset at an average interval of 8.9 days, though the timing is often unannounced and varies based on infrastructure stability.
• Banked Reset Feature: A new 'banked reset' system allows users to manually replenish their weekly usage limits, with functionality recently expanded to web and mobile platforms.
• GPT-5.6 Sol Accessibility: The removal of the 5-hour rate limit is specifically designed to allow users to explore the boundaries of the GPT-5.6 Sol model.
• Infrastructure Resilience: Resets are frequently tied to the team's efforts to maintain 100% weekly usage limits while scaling infrastructure at 'lightspeed.'

In-Depth Analysis

Infrastructure Scaling and User Milestones

The data provided by the Codex Resets tracker paints a picture of a platform experiencing explosive growth. The announcements logged by @thsottiaux reveal a rapid progression in active user counts, moving from 7 million to 8 million, and most recently hitting the 9 million mark. This growth trajectory places immense pressure on OpenAI's infrastructure. The 'out of the blue' nature of usage resets suggests that these replenishments are closely tied to the team's ability to scale their systems in real-time.

As the team 'iterates at lightspeed,' the resets serve as a relief valve for a user base that is expanding faster than ever. The mention of hitting 9 million users 'way earlier' than expected on a single day indicates that user engagement is outpacing internal projections. To maintain reliability, the team has had to balance the rollout of resets with the 'millions of things' required to keep the systems up. The goal remains to provide a 'sweet 100% weekly usage limit' to all paid users, ensuring that the service remains dependable despite the massive influx of traffic.

The Evolution of Usage Management: Banked Resets

A significant shift in OpenAI's approach to usage limits is the introduction of the 'banked reset.' Previously, usage limits were largely managed through automatic system-wide resets. However, the introduction of banked resets—initially limited to the desktop app and now available on web and mobile—gives users more agency over their quotas. This feature allows users to celebrate milestones, such as the 7M user mark, by applying a reset to their own accounts when needed.

This transition was not without technical hurdles. The logs indicate that while rolling out the banked reset feature to 500,000 users of ChatGPT Work and Codex, an issue affected approximately 10% of those who attempted to use it. The expansion of this feature to web and mobile platforms suggests a move toward a more flexible, user-centric model of resource allocation. By 'banking' a reset, users can replenish their weekly usage at their convenience, which may help smooth out the spikes in demand that occur during global, system-wide resets.

Strategic Access to GPT-5.6 Sol

The most recent updates highlight a specific focus on the GPT-5.6 Sol model. By removing the 5-hour rate limit, OpenAI is explicitly encouraging users to 'explore the boundaries' of this specific model. This move suggests that the infrastructure has reached a level of stability where the previous restrictive rate limits are no longer necessary to prevent system failure.

The removal of these limits is framed as an invitation for users to discover 'how ambitious' they can be with the AI. This suggests that GPT-5.6 Sol is positioned for high-intensity, complex tasks that were previously hindered by frequent rate-limiting. The tracker shows that these resets and limit removals are often timed to coincide with weekends, potentially to allow professional users and developers more freedom to experiment during off-peak or dedicated project hours.

Industry Impact

The patterns observed in the Codex Resets tracker have broader implications for the AI industry, particularly regarding transparency and user retention. In an era where high-performance compute is a scarce resource, OpenAI's strategy of 'surprise' resets and milestone-based rewards (like banked resets) serves as a powerful tool for maintaining user engagement.

Furthermore, the data regarding reset intervals—ranging from a frequent 8.9-day average to a long gap of 67.7 days—highlights the volatility of maintaining large-scale AI services. For the industry, this underscores the importance of infrastructure resilience. As OpenAI scales toward 10 million users and beyond, the ability to provide consistent access to advanced models like GPT-5.6 Sol will be a key differentiator. The shift toward user-controlled 'banked' resets may also set a new standard for how AI providers manage limited resources, moving away from rigid, scheduled resets toward a more dynamic, credit-based system.

Frequently Asked Questions

Question: What is a "banked reset" and how can I use it?

Answer: A banked reset is a credit added to a user's account that allows them to manually replenish their weekly usage limits for Codex and ChatGPT Work. Originally available only on the desktop app, it can now be applied through both the web interface and mobile applications.

Question: How often does OpenAI reset the usage limits for Codex?

Answer: According to the Codex Resets tracker, the average interval between resets is approximately 8.9 days. However, this is not a fixed schedule; the longest recorded gap between resets was 67.7 days, and resets are often announced spontaneously based on system milestones and infrastructure capacity.

Question: Is the 5-hour rate limit still in effect for GPT-5.6 Sol?

Answer: Recent announcements indicate that the 5-hour rate limit has been removed for Codex and ChatGPT Work users. This change is intended to allow users to explore the capabilities of the GPT-5.6 Sol model more freely and ambitiously without being interrupted by short-term usage caps.