Bolt.new / AI App Builder
Bolt adds Forge and a cheap tier
Until now Bolt only offered its own in-house agent, with no budget tier. On Sept 14, it launched a new open-model agent, "Bolt Forge," alongside a $9/month "Bolt Lite" plan.
Beyond a single in-house agent
Cheap-model options are already table stakes among AI app builders — Bolt is catching up.
StackBlitz's Bolt.new had shipped only two in-house agents, Standard and Max, effectively at a single price tier. According to a press release dated September 14, 2026, the company added a third agent, Bolt Forge, that runs not on Bolt's own models but on open-weight ones: GLM 5.3, GLM 5.3 Flash, Kimi K3, and DeepSeek v4 Pro.
The rationale is laid out in the official blog post "What is Bolt Forge?" Because open models are cheaper for Bolt to run, existing Pro-and-up subscribers get Forge bundled in at 50x the usage allowance of Standard mode, at no extra cost. Letting people build a lot more, cheaply, is a clear play to retain users who might otherwise have churned over price.
Forge and Lite, in figures
The generous terms come with a catch. Per the Bolt Lite launch page, the 50x allowance is a special condition tied to the "research preview" running September 14 to October 14, 2026. Build sessions are anonymized and, for opted-in users, shared with open-model lab Arcee, feeding a trillion-parameter-class open-weight model whose training run is slated to start in October. In effect, users trade session data for cheap or free usage.
Three ways into Forge
Already Pro or above
Existing Bolt Pro and Team subscribers see Forge appear in the agent picker automatically, with no extra charge and no setting to flip.
Sign up for Bolt Lite
If Forge alone is enough and you don't need Standard or Max, the $9/month Bolt Lite plan is the cheapest entry point. Sign up before Oct 14 to keep the $9 price going forward.
Pick a model
Inside Forge you can still switch between GLM 5.3, GLM 5.3 Flash, Kimi K3, and DeepSeek v4 Pro. DeepSeek v4 Pro is flagged as experimental and its behavior may still change.
Who this is for
Solo engineers
Side-project and indie builders watching fixed monthly costs get most Standard-grade work covered by the $9 Bolt Lite plan running on Forge.
Designers
During heavy prototyping, Forge's wider allowance means more iterations. For final-polish work, switching back to Max still makes sense.
Business / budget owners
Teams already on Pro or above get Forge for free. Teams evaluating Bolt for the first time can pilot it cheaply through Bolt Lite before committing further.
| Standard / Max (existing) | Forge (new) |
|---|---|
| Bolt's own models | GLM 5.3 family, Kimi K3, DeepSeek v4 Pro |
| Standard usage cap | 50x Standard during the research preview |
| No standalone plan | Bolt Lite ($9/mo) offers Forge on its own |
| Session data handled normally | Opt-in sessions feed Arcee AI's training data |
Behind the low price, your build sessions become training material for the next model.
Good news if you're watching spend
For developers and designers already on Pro or above, Forge is effectively a feature expansion at no extra cost. Since it's bundled in free, the sensible move is to push lighter prototyping work onto Forge and save Standard/Max allowance for tasks that genuinely need the extra precision.
The $9/month Bolt Lite price, though, is explicitly tied to the September 14–October 14, 2026 research-preview window. Nothing in the current announcement commits to keeping that price or the 50x allowance permanent once signups close on October 14. Budget owners should plan to check whether the terms hold once the preview ends. Individual developers, meanwhile, have a clear one-month window to sign up for Bolt Lite now and test whether Forge's quality is good enough for their actual workload before the October 14 deadline.