Vibe coding platform Base44 launches own model as AI startups seek defensibility

Base44, the innovative vibe coding platform snapped up by Wix for $80 million just a year ago, is making headlines with the launch of its own AI model aimed at simplifying app creation through natural language processing. This strategic development comes at a critical time when the AI community is increasingly questioning the suitability of frontier models for diverse applications and the long-term sustainability of businesses relying on external AI frameworks. Located in the Bay Area, Base44’s new initiative highlights its ambition to not only compete but potentially surpass existing frontier models. The company's founder, Maor Shlomo, emphasized that by training and owning their own large language model (LLM), Base44 can optimize various operational aspects, including latency, cost, and efficiency. This move is particularly noteworthy as it positions Base44 against competitors like Lovable, a Swedish startup that reached unicorn status while depending on third-party models. Despite the competitive landscape, Shlomo anticipates that other significant players in the market will also embark on training their own models as they scale and gather sufficient data. Jonathan Userovici, a general partner at VC firm Headline, which supports various AI startups, pointed out that data, along with distribution and technology stack, is crucial for establishing a defensible position in the AI sector. As a result, companies with strong brand identities are increasingly leveraging their data and infrastructure to enhance their competitive edge, a trend that Base44 is keenly following. The company has revealed that its inaugural LLM, dubbed Base1, was developed using a dataset derived from tens of millions of genuine user interactions on its platform. This dataset is set to expand as Base44 grows, even as competition intensifies from both vibe-coding startups and leading AI labs like Cursor and Grok, which are now under SpaceX. Shlomo believes that despite the advancements in general models, specialized solutions like Base44’s can offer distinct advantages in targeted applications. Userovici, however, warns against underestimating frontier models, citing the experience of legal tech startup Harvey, which opted against developing its own model. He argues that while most applied AI companies may not transition into frontier labs, Base44's decision reflects a broader market shift towards optimizing inference costs, a concern that enterprise customers are increasingly vocal about. As businesses seek to manage their AI expenses while maintaining performance, Base44’s initiative could prove vital. Shlomo noted that the goal is to create a model that aligns more closely with user needs, ultimately providing a faster and cheaper alternative to existing frontier models. Even though the financial implications for Base44 remain a bit uncertain, the company claims that owning its model enables better control over computing and inference costs, potentially leading to improved margins over time. This could be a significant advantage, especially as Base44's parent company recently announced a 20% workforce reduction. In contrast, Base44 has continued to expand its team and recently celebrated surpassing $100 million in annual recurring revenue, though still trailing behind Lovable, which reported $500 million in ARR. Shlomo is optimistic that the substantial engineering investment in developing Base1 will solidify Base44’s position as the only vertically integrated vibe-coding application, a unique space where it controls its distribution, data, and infrastructure all at once.