Models & rates
The Agent and the AI steps in Loops run on the models below. Every plan includes a set amount of AI usage each month (the free trial, one amount for the whole trial); when it runs out, you can keep going with on-demand usage, billed monthly after use, up to the limit you set. See Billing.
Usage is counted in tokens, the pieces of text a model reads and writes. Each request is charged at the model's rates, per million tokens, in US dollars:
- Input: what the model reads: your message, the chat so far, and what the Agent looked up. Images the Agent looks at, such as your media and your ads' creatives, count as input tokens too.
- Cache write and cache read: input the model keeps for the next steps of the same task. Reading it back costs much less than reading it fresh.
- Output: what the model writes, including its reasoning.
Anthropic
| Model | Input | Cache write | Cache read | Output |
|---|---|---|---|---|
| Claude Opus 5.5 | $5.28 | $6.60 | $0.264 | $26.40 |
Web searches: $12.00 per 1,000.
Rates are per million tokens and exclude applicable taxes. Your organization's Usage page in Settings shows every request with its model, tokens and charge, and lists these rates under Models & rates. MCP and API access with your own AI tools is included on every plan and isn't counted as usage.