🧠 Kimi K2.7 and Fable 5
Two new models are now available from the model picker: Claude Fable 5 (Anthropic's most capable model to date) and Kimi K2.7 (an open-source model, best known for its lower token costs).
Kimi K2.7 is an open-source model that runs at a fraction of the cost of the main frontier providers, and we host it ourselves on US-based infrastructure. It's not the one to reach for on your hardest questions, but for quick syntax help, simple transformations, and well-defined lookups, and gets you there cheaper.
One of our engineers benchmarked it against Opus, so you don't have to. Read how Kimi stacks up in our blog
Claude Fable 5, Anthropic's most capable model to date, is also back. One thing to know: Fable requires a data retention opt-in for safety monitoring, so admins need to enable model data retention before it shows up for your team. We recommend reviewing with your security and legal teams first.
☑️ Admins can set the default model
Admins can now choose which model the agent defaults to for the whole workspace, so every new thread starts on the right model instead of leaving it up to each person.
🎨 Brand your generative apps
Every app your team ships should look like it's actually yours.
Set your colors, typography, spacing, and style rules once in a design.md file in your workspace guides, and the Hex agent uses it for every generative app your workspace builds from then on.
A one-off prompt can still override it, and you don't have to write it by hand. Give an agent your existing brand guidelines and it'll draft the file for you.
🛠️ Other improvements
- Import and export generative apps as YAML: Generative app code can now be exported and imported as YAML, so you can share, template, or version an app just like the rest of your Hex project.
- Faster notebooks and apps: Chart and pivot cells now run in parallel instead of one at a time, so notebooks and apps with a lot of visual cells finish running faster. App runs also now prioritize whichever tab you're actively looking at.
- Easier SSO login: Workspaces using SSO now have a working "Log in with SSO" option right on the main login page, including for MCP logins. No more getting stuck guessing which login path to use.