OpenAI's upcoming chat-and-work merger in ChatGPT signals the end of all-you-can-prompt pricing

Your ChatGPT subscription now has a usage cap that affects more than just chat.

OpenAI is merging its ChatGPT and Codex services, impacting its pricing structure starting in December. Users on Plus, Pro, and Business plans will now share a single token budget across both services, meaning complex tasks and models will consume this budget more quickly. While chats with "Instant" models will not count against the budget, heavy users on lower-tier plans may experience reduced output quality. Users can monitor their token usage, and if they exceed their limit, they can wait for a reset, upgrade their plan, or purchase additional credits. This integration, announced on an updated OpenAI support page, follows OpenAI's earlier plans to combine Chat and Work features and mirrors Anthropic's similar integration of Chat and "Cowork" in its Claude product. The move appears to be a response to growing complexity and user feedback regarding ChatGPT.

This consolidation signals a shift away from unlimited or "all-you-can-prompt" pricing models in AI assistants, pushing towards metered usage that aligns costs with computational resources consumed. The shared token budget across different functionalities, including coding and general chat, means users must now carefully manage their interactions to avoid exceeding limits, potentially impacting workflow efficiency for those with diverse AI needs. Companies are likely to face increased scrutiny over their AI spending as token consumption becomes a more tangible and capped expense.

The move by OpenAI, alongside Anthropic's prior integration, suggests a market trend towards streamlining AI interfaces while simultaneously implementing more granular cost controls. This could lead to a tiered AI service market where advanced capabilities or higher usage volumes necessitate more expensive plans or custom enterprise solutions. Users may need to re-evaluate their AI tool strategies, considering the trade-offs between cost, performance, and the new token-based limitations.