OpenAI’s always-on agents are free, until one specific thing happens

OpenAI recently introduced "Dots," a set of always-on AI agents that operate continuously without counting against a user's standard usage allowance during the launch period. According to OpenAI's head of core products and platform, Thibault "Tibo" Sottiaux, the primary Dot included in a subscription plan is available 24/7 and does not consume any of the plan's allowance while directly working on tasks.

However, when a Dot is asked to create a task in another OpenAI product like Codex, the Codex task will draw usage as usual.

OpenAI plans to allow users to upgrade the speed and bandwidth of their Dots in the future, but this additional capacity will cost money while the baseline functionality remains free. Currently, OpenAI expects to have several million Dots online within days of the launch. The company absorbs the baseline activity costs during the launch period for each user's primary Dot. Once the initial launch period ends, the included allowance for Dots usage will be clearly defined per plan.

The way a Dot processes work can directly impact the user's costs, as it can choose to complete tasks itself, call included tools, or send tasks to metered products like Codex. Currently, OpenAI is absorbing the usage costs for Dots during the launch period, but they expect the cost to be absorbed by users after the launch period ends. Users may not be notified before their Dot transitions from free usage into metered usage, potentially leading to unexpected charges.