Pro, Pro+ or Ultra: which plan should you pick?

The short answer

• Pro if you mostly use Tab and chat, with a handful of agent runs a day.
• Pro+ if you use Agent for real work every day. It becomes the cheaper option once you use more than $60 of models a month.
• Ultra if you run agents most of the day, or several in parallel. It becomes the cheaper option above $210 a month.

How the plans work

Each plan is a subscription that includes a monthly amount of model usage, measured in dollars at each provider's list price. Beyond that, usage is billed on demand at the same list price, up to the hard cap you set.

The break-even points

Your total monthly bill is the plan price plus any usage above what the plan includes. That gives two simple break-even points:

• Pro vs Pro+ at $60 of usage. On Pro, $60 of usage costs $20 + $40 = $60, the same as Pro+. Above that, Pro+ is cheaper.
• Pro+ vs Ultra at $210 of usage. On Pro+, $210 of usage costs $60 + $140 = $200, the same as Ultra. Above that, Ultra is cheaper.

Estimating your usage

A typical agent request reads around 8,000 tokens and writes around 1,000. Depending on the model, that costs somewhere between half a cent and seven cents. Multiply by how many requests you make in a working day, then by about 21 working days:

The estimator on the pricing page does this calculation for you and highlights the cheapest plan.

Reasons beyond price

• Priority capacity. Pro+ gets priority at peak hours, and Ultra gets the highest priority, even on the day a new model launches.
• Max context. Pro+ and Ultra unlock the largest context windows each model offers.
• Parallel agents. Ultra removes the limit on how many background agents can run at once.