Bolt Adds DeepSeek V4.1 Flash to Its Forge Preview
Bolt has integrated the DeepSeek V4.1 Flash model into its Bolt Forge research preview, offering developers a highly cost-effective option optimized for rapid visual design iterations.

Bolt has expanded its Bolt Forge research preview by adding DeepSeek V4.1 Flash to its agent picker. This new addition provides a significantly cheaper alternative to DeepSeek V4 Pro, consuming roughly one-tenth of the usage cost. This dramatic price reduction allows developers to get about 10 times more usage under the same allowance, making it highly efficient for repetitive design tasks like tweaking layouts, typography, and responsive structures. Beta testers have already favored this model for its visual output on landing pages and marketing websites.
The addition of DeepSeek V4.1 Flash expands the Forge lineup, which already features GLM 5.3 Flash, GLM 5.3, Kimi K3, and DeepSeek V4 Pro. Collectively, these open-weight models achieved a score of 92.2 on Bolt's internal Build Index. This is highly competitive with the 101.0 score registered by the platform's top-performing closed, paid model. For practitioners, this means they can access open-source capabilities that deliver roughly 91 percent of the performance of proprietary alternatives at a fraction of the cost.
To encourage adoption, Bolt is offering Pro subscribers up to 50 times more free usage through October 14. The company has also introduced Bolt Lite, a new $9 monthly tier aimed at students and side-project developers that provides exclusive access to Forge. However, accessing these expanded allowances requires users to opt into data sharing. Under this agreement, Bolt anonymizes and shares build-session data—including prompts, code, errors, and manual edits—with Arcee AI to help train a planned trillion-parameter open-weight model.
For developers, this update introduces a more flexible workflow where they can use the cheaper DeepSeek V4.1 Flash for rapid prototyping and visual polish, then switch to heavier models for logic-heavy builds. Because Forge remains an experimental preview, Bolt advises developers to duplicate their projects before switching models and to use Standard or Max modes for complex production environments.
This is our own summary of reporting by AlphaSignal



