OpenAI plans to run a small test that merges ChatGPT Chat and Work agent features into one interface. The company wants users to move from asking a quick question to finishing a larger project without changing screens. An email seen by BleepingComputer says the first trial will include a limited group of users. OpenAI has not said how broad the test will be, or when it might reach general availability. That raises a practical question: how much of a user’s allowance will longer agent tasks consume?
One chat, two kinds of work
OpenAI currently frames Chat as the place for quick questions and brainstorming. Work handles longer jobs such as research, spreadsheets, reports, presentations and published Sites. People can upload files, use connected plugins and guide the agent as it works. The proposed merger would fold those steps into a single conversation. Codex would stay separate.
Anthropic moved first
Anthropic announced on September 16 that Cowork and chat are merging into one Claude. The combined experience started with Pro, Max and Ultra subscribers. Enterprise followed on October 23, while Team and Free plans are still pending. Anthropic also introduced Claude Docs and Slides for editing documents and presentations. ChatGPT Work has similar features, including Sites publishing. Both companies are pushing users toward task completion, not just answers.
Costs draw more attention
The pricing details may matter as much as the interface. OpenAI already shares an allowance between Work and Codex, and users can buy credits after they run out. On the Pro $500 plan, Ultrafast uses included usage at eight times the standard rate, then draws credits at six times the standard rate. In December, Plus, Pro and Business subscribers will share a token-based allowance across ChatGPT and Codex. Instant chat will not count toward that limit. Users will be able to check remaining usage and reset times under Settings > Usage.
The change follows OpenAI’s October 29 decision to double Work and Codex allowances for Pro $100 and $500 subscribers. Some Pro $200 accounts also received higher limits. Even so, Codex users have complained that allowances seemed to drain faster than expected. OpenAI executive Thibault Sottiaux disputed claims of undisclosed cuts. Kingy AI first called the claims “supported but unproven,” then shifted to “partly confirmed; blanket cut unproven” after acknowledging cache-efficiency problems.
Adoption is still uneven. Gartner expects spending on AI agents and assistants to rise from $29.2 billion in 2026 to $65.5 billion in 2027. An IDC study of 1,900 organizations found only 3.1% at the highest AI-maturity stage, while 61.3% were in the two earliest stages. For OpenAI, predictable costs may decide whether the merged interface feels useful or just expensive.
