Correction, October 9, 2026: Bolt’s official announcement and preview launch are dated September 14, not September 11. European Economic Area, UK and Swiss sessions are exempt from training.
What happened
Bolt.new announced Bolt Forge, a new agent inside its AI app builder. Bolt dates the announcement and research-preview launch to September 14, 2026. It’s the browser-based tool where you describe an app in plain English and it writes and runs the code for you. Its Visual Edits toolbar changed how users edit an app. Forge adds a model choice with consequences for how sessions are used.
Forge sits in the agent picker next to the Standard and Max agents Bolt users already have. The difference is what it runs on. Standard and Max use Bolt’s regular paid models. Forge runs only on open-weight models, meaning models whose underlying trained files are published for anyone to use or build on, rather than kept private by one company. By default Forge uses a model called GLM 5.3 Flash, with GLM 5.3 available alongside it, and Kimi K3 and DeepSeek v4 Pro offered as experimental options.
The trade Bolt is offering: every individual Pro plan gets up to 50 times more Forge usage than normal, at no extra cost, through October 14, 2026. In exchange, Bolt asks you to let your Forge sessions help train future open-weight models, through a partnership with Arcee AI, a company that builds open AI models. Bolt says this training is always opt-in: a consent screen explains the trade before you build, and sessions are anonymized with secrets stripped before they’re used. Bolt says this sharing applies only to Forge, leaving Standard and Max unchanged.
European Economic Area, UK and Swiss sessions are not used for training or datasets. Team and Enterprise workspaces are excluded from Forge.
On Bolt’s own quality benchmark, called the Bolt Build Index, Forge’s open models score 92.2 against 101.0 for Bolt’s top paid model, roughly 91% of the top score. That ratio is not a percentage of app quality.
Why it matters
For eligible sessions outside the exempt regions, the allowance comes with a data-sharing decision. Sessions you consent to share become training material for models other companies and developers can use. Check whether your business logic, product ideas or client work can be used that way before consenting.
The benchmark measures performance on Bolt’s test projects. A lower score does not establish how often Forge will fail on your own app.
Bolt’s offer combines more usage with an opt-in dataset program.
Who should care
This matters most if you’re on a Bolt Pro plan and regularly bump against your usage limit, since Forge is aimed squarely at that friction. It’s also worth watching if you’re curious how open-weight models compare to paid ones for real app-building work, since results and usage on your own project may differ from Bolt’s benchmark.
If the data-sharing terms do not fit your project, declining consent keeps you in Standard or Max. Builders in exempt regions receive the Forge allowance without sharing sessions for training.
What builders should do next
Read the consent screen and check whether your region and workspace qualify before deciding what to put into a Forge session. Then try one routine task in Forge and review the result and usage it consumes.
End of article