Product
Faster public pages and steadier site builds
Public pages such as the home page, pricing, docs and the Newsroom are now served from a cache close to you, so they load faster.
Adds edge caching for public pages, asset versioning, and clearer errors when reasoning models exceed token limits.
Public pages are the same for every visitor, so we now let them be cached for a few minutes. A new article or a site update can take up to five minutes to appear.
Stylesheets and scripts now carry a version in their address. Your browser can keep them until the next release instead of checking again every few hours. The main fonts start loading together with the page.
Nothing changes once you are signed in. Your workspace, account pages and everything after login are never cached.
Site builds and structured answers
Reasoning models sometimes used their output allowance on thinking and returned no answer. Structured requests now ask for more output room when the model permits it. If an answer is still cut short, the error says it reached the output limit.
Some custom page sections failed the build checks when their styles used CSS variables. The generated project now accepts these valid style properties.