Bolt Forge launches with open AI models, GLM 5.3 Flash leads usage at 46%
Bolt released Forge, an AI agent for its browser-based app builder that exclusively runs open-weight models including GLM 5.3, GLM 5.3 Flash, Kimi K3, and DeepSeek v4 Pro. No API keys or extra setup are required. In the first 48 hours, GLM 5.3 Flash captured 46% of prompts, with users favoring its speed and cost for larger tasks. The platform offers free usage until October 14 or a $9 Bolt Lite plan.
Editorial responsibility
- No named human review is recorded for this page.
- Reports are grouped by semantic similarity and deterministic rules. Language models may assist titles, summaries, translation and cross-source analysis; the page reads the event directly, while its address stays stable when the title changes.
- Summary covers the current reports
Cross-source coverage
Reporting timeline
KIMI K3. DEEPSEEK V4 PRO. GLM 5.3. GLM 5.3 FLASH. ALL AVAILABLE DIRECTLY INSIDE BOLT FORGE. NO API KEYS. NO EXTRA SETUP. JUST PICK A MODEL AND BUILD!! https://t.co/oU8Boy0pOL
A post from DataChaz on X announces that four AI models — KIMI K3, DeepSeek V4 Pro, GLM 5.3, and GLM 5.3 Flash — are now available directly inside the Bolt Forge platform. The announcement emphasizes that no API keys or extra setup are needed; users can simply pick a model and start building. This integration lowers the barrier for developers and users to access these models for their projects.
Read sourceOpen models face distribution friction, Bolt Forge data shows GLM 5.3 Flash leads usage
An analysis by Rohan Paul argues that open AI models have a distribution problem: making weights or an API available online does not integrate them into a developer's workflow. The post cites @boltdotnew Forge as an example where this friction is eliminated, with four open models available in a single picker used by over 11 million builders. According to the data, GLM 5.3 Flash accounts for 54% of Bolt Forge prompts to date, while DeepSeek V4 Pro, GLM 5.3, and Kimi K3 share the remaining usage. The analysis suggests that within a build loop, model capability competes against latency, cost, and the inference budget provided by the product. A theoretically stronger benchmark model may be less appealing if every iteration is slower or more expensive, indicating that practical factors heavily influence model adoption in development workflows.
Read sourceBolt Forge usage data: GLM 5.3 Flash leads with 46% share in first 48 hours
Bolt has published the first 48 hours of usage data from Bolt Forge, a mode within its AI app builder that allows users to run open-source models. The data reveals a breakdown of model usage: GLM 5.3 Flash captured 46% of prompts, followed by DeepSeek V4 Pro at 21%, GLM 5.3 at 17%, and Kimi K3 at 16%. Notably, GLM 5.3 Flash, described as the fastest and cheapest of the four models, not only captured the largest share of prompts but also received prompts that were, on average, twice as large as those sent to the other models. This suggests users favored the model for its speed and cost efficiency for larger tasks.
Show 2 older updatesHide older updates
Bolt.new launches Forge mode with DeepSeek v4 Pro and GLM 5.3 Flash models
Bolt.new (@boltdotnew) has shipped a new mode called Forge, which integrates open AI models directly into its web builder. The platform has made DeepSeek v4 Pro and GLM 5.3 Flash available for testing. Users who opt in to model training receive an unusually large amount of usage. The offer is free until October 14, or users can lock in the Bolt Lite plan for $9. This launch is positioned as a significant development for 'vibe-coders,' a community of developers who build with AI assistance.
Read sourceBolt releases Forge, an AI agent for browser-based app building using open-weight models
Bolt has released Forge, a new AI agent within its browser-based app builder that exclusively runs on open-weight models. The default lineup includes GLM 5.3 and GLM 5.3 Flash, with Kimi K3 and DeepSeek v4 Pro available as experimental options. All models are accessible under a single monthly usage bar with no daily limits. On the Bolt Build Index, the company's proprietary benchmark for project completion within the builder, the Forge models scored 92.2, compared with 101.0 for Claude Opus 5, indicating a performance gap against the leading proprietary model.