Frequently asked questions.

Yes. Hornbeam Free gives limited access to Thicket, including our current models, chat, search and creation, and Thicket for code for agentic coding in your terminal.

Catkin is the fastest and cheapest, and handles classification, extraction and short drafting well. Bract is the balanced default for most product work. Burl is the reasoning model — reach for it on long context, multi-step agents and hard code.

Per token, billed separately for input and output. Catkin runs $0.14 and $0.42 per million tokens; Burl runs $1.85 and $6.30. Cached input is billed at 12% of the input rate, and there is no charge for failed requests.

The console breaks spend down by key, model and day, and exports to CSV. You can set a monthly cap plus alert thresholds per workspace, and usage is also readable from the billing endpoint if you would rather graph it yourself.

Free covers everyday use with lower rate limits and a shorter context window. Pro is $19 a month: priority capacity, the full 256k context, unlimited project files, and early builds of new models a few weeks before general release.

There are rolling limits rather than hard monthly quotas, and they reset every five hours. Free sits at roughly 40 messages per window, Pro at around 12 times that. Heavy code sessions and long attachments draw down the window faster.

Catkin and Bract ship with open weights you can run on your own hardware or in your own cloud account. Burl is available for on-premise deployment under a commercial licence, with support for air-gapped installs.

Yes. Students and teaching staff at an accredited institution get Pro at 60% off for twelve months, renewable while you are enrolled. Verify with your institutional address and the discount applies at the next invoice.

Enterprise adds SSO and SCIM, region-pinned inference, a signed DPA, SOC 2 Type II reporting and a named solutions engineer. Support runs to a 30-minute response on severity-one issues. Talk to the team to scope a deployment.