Release Notes

Release Notes

Every Core dna product, release by release, newest first. Each product ships on its own schedule.

Platform 2026.4.2

What’s new in Core dna Platform 2026.4.2

Performance and Reliability Improvements

Records with long revision histories open and save reliably, and sites can serve more file types as static assets. Email addresses are checked before sending.

Key highlights

  • Long revision histories open and save reliably.
  • More static file types: a site can serve extra file types as static assets.
  • GPT-4o by default for the AI content tools.

Changes in behaviour

  • AI model: a site still set to the retired GPT-4 Turbo model should choose a supported model in the AI Model setting.

Long revision histories

Records with very long revision histories, such as pages, blog posts, products, forms, FAQ entries and files, now open and save reliably. Past revisions are loaded only when they are needed, so heavily edited content stays fast.

More static file types

A site can now serve extra file types as static assets by adding a .static-assets.json file at the site root, listing each extension with its content type. Executable code, templates, dependency files and hidden files are never served, whatever the file says.

JSON
{
    "allowedTypes": {
        "glb": "model/gltf-binary"
    }
}

GPT-4o by default

The AI content tools now use OpenAI's GPT-4o by default, and the retired GPT-4 Turbo option is gone from the AI Model setting in the Global settings.

Other improvements and bug fixes

  • Sites can now extend the file types served as static assets by adding a .static-assets.json file, allowing non-standard asset types to be delivered directly. Built-in safeguards ensure executable code, source templates, dependency files and hidden files are never served, even if misconfigured.
  • Email addresses are now trimmed and validated before sending, so messages with stray whitespace or malformed addresses no longer silently fail to deliver. Delivery failures are now reliably recorded as failed instead of being marked as sent, and reply-to display names are preserved on outgoing mail.
  • The AI content tools now use OpenAI's GPT-4o model by default, and the discontinued "GPT4 Turbo" option has been removed from the model selector to prevent failures from the retired model.
  • Fixed an issue where updating the metadata of an image revision could fail with a confusing error. Requests for a revision that no longer exists now return a clear "not found" response.
Go deeper

Was this page helpful?

Your answer goes to the docs team.