OpenAI to merge ChatGPT and Work into a single usage pool in December

The transition to a shared pool

OpenAI is preparing to change how it meters its most powerful features for paid subscribers. Starting in December, the company plans to merge its 'Chat' and 'Work' surfaces into a single experience. This means that instead of having separate limits for different types of tasks, users on Plus, Pro, and Team plans will share one bucket of usage.

Currently, the system treats different features with distinct caps and reset clocks. For example, image generation, voice, and file uploads often sit in their own categories. Under the new system, a user who generates a product mockup will be drawing from the same allowance as someone running a long, autonomous coding session.

This shift was first signaled during a podcast appearance by Tibo Sottiaux, who oversees ChatGPT and Codex at OpenAI. While he did not provide a specific date during the interview, a checkout notice for some subscribers now confirms the December timeline. The change represents a move toward a more unified way of managing the high costs of AI compute.

What stays outside the meter

Not every interaction with ChatGPT will be affected by this new shared limit. OpenAI has stated in its help documentation that plain chat on the fast, default model remains outside of the metered pool. This means that simple, everyday questions that do not require heavy processing will still be accessible without depleting the shared allowance.

The new pool is specifically designed for tasks that lean on heavier compute. This includes 'Work'—the slower, agentic mode built for multi-step jobs like research—and 'Codex,' which handles complex coding tasks. Because these features require more power, they are being grouped together to manage how much of the company's resources they consume.

The distinction is important for how users will need to plan their workflows. A student summarizing a reading list will likely not touch the new pool at all. However, a startup founder who uses the AI as an all-day collaborator for drafting decks, debugging code, and generating logos will hit a shared ceiling that did not exist in the previous model.

A history of usage friction

This consolidation comes after a period of technical and logistical hurdles for OpenAI. When ChatGPT Work first launched this summer, the company struggled with usage limits. They reportedly had to reset the limits for Codex and Work users twice in a single day because the highest compute settings were being triggered too easily by accident.

OpenAI admitted it did not get everything right during that initial rollout. To address the resulting backlash, the company loosened limits and provided Codex and Work subscribers with up to 50% more usage in August. These repeated adjustments suggest that the company is still refining how it balances user demand with the underlying costs of running large models.

The move to a shared pool is a way to create a more predictable system for both the company and the user. By grouping all high-intensity tasks together, OpenAI can more accurately track the 'burn' of tokens. However, the immediate reaction from the community has been wary, with many users expressing concern over how the specific counting will work once the merger goes live.

Industry-wide consolidation

OpenAI is not the only company moving toward this unified model of AI interaction. Microsoft recently confirmed it is merging its consumer and enterprise Copilot apps into a single application. This move folds chat, code, and autonomous agents into one surface, mirroring the structure OpenAI is proposing for its own products.

Other major players are following similar paths to streamline the user experience. For instance, Anthropic has already folded its 'Cowork' feature into its main Chat interface. These moves suggest that the industry is moving away from separate products for different tasks in favor of all-in-one platforms that handle everything from simple queries to complex agentic work.

This trend highlights the reality of compute as the primary constraint in the AI market. As companies like Cursor and GitHub Copilot shift toward consumption-based billing or workplace platform models, metering becomes the primary tool for pricing. For the reader, this means the era of 'unlimited' feel for specific features is being replaced by a more calculated, shared economy of AI usage.

Impact on teams and founders

The most significant impact of this change will be felt by teams and founders who use AI for diverse professional tasks. Under the old system, a marketing team could generate images without worrying about the limits of the engineering team's coding tools. In the new system, these two departments will be competing for the same monthly allowance.

Companies paying for seats per employee will need to develop new strategies for internal budgeting. A graphic generated by a marketer and a code refactor performed by an engineer will now draw from the same bucket. This requires a level of coordination that wasn't necessary when the limits were siloed by category.

As the December deadline approaches, power users should begin to monitor their current usage patterns. Understanding which tasks fall into the 'Work' or 'Codex' categories will be essential for avoiding unexpected caps. The shift marks a transition from a feature-based limit to a compute-based limit, reflecting the true costs of running high-end AI agents.