Events, as they come.
Launches, giveaways, free models and maintenance windows — anything the workspace is running, with the dates and the details that matter.
Nothing live yet.
How events work
Free models are the part people ask about most, so the rules are written out here rather than buried in a support reply.
Two kinds. A model event frees one specific model for a while, drawing on a token budget. An announcement is anything else the workspace wants to publish — a launch, a maintenance window — and has no effect on requests or prices.
Call it the way you call any other model, on the same endpoint, with the same API key. Nothing changes in your code. While the event is running, requests to that model are not billed to your credits. Switch the model id and you are done.
No. Every model event carries a token budget. Usage is counted as you go, and once the budget is used up the event is exhausted and the model goes back to its listed price. The progress bar on a card shows how much is left.
The provider running the model reports them, and we meter what comes back. That is also why the budget is tracked in tokens rather than credits — credits are our unit, tokens are the provider's.
They stay free. A request is judged by whether the event was running when it was made, so the event ending mid-stream does not retroactively bill you. The tokens it used do come out of the budget.
Because the model itself stopped being available. A giveaway for a model nobody can reach would be a promise we cannot keep, so an event is hidden and its model is not flagged free whenever the model is not active. The event is not deleted — it returns if the model does.