Base44 launches proprietary AI model to compete with frontier labs in vibe coding

Base44, the natural-language app-building platform that Wix acquired for $80 million last year when it was barely six months old, has begun rolling out its own in-house AI model. The move underscores a broader tension in the AI startup ecosystem: whether companies built atop third-party foundational models can remain competitive long-term.

Why the shift matters#

The question of defensibility has become urgent for AI companies riding the wave of generalist models. While early vibe-coding platforms like Base44 thrived by integrating tools such as Claude Opus, the calculus is shifting. As frontier AI providers — including OpenAI, Anthropic, and xAI — move into specialized domains and launch their own code-generation features, dependent startups face margin pressure and loss of differentiation.

Base44's founder, Maor Shlomo, framed the decision in terms of optimization: "training and owning the model as part of [our] entire stack allows us a lot more optimizations on latency, cost, and efficiency." The company's new model, called Base1, was trained on tens of millions of user interactions collected across the platform over time.

Defensibility through vertical integration#

Industry analysts see three pillars of defensibility for AI startups: data, distribution, and infrastructure. Base44 is betting on owning all three. Unlike competitors such as Lovable — which reached unicorn status by relying on external models — Base44 is building what Shlomo calls "vertical integration," controlling the full stack from user feedback loops to model training to inference.

The move also addresses cost pressures. Enterprise customers increasingly question the return on investment from using cutting-edge models for every task. Shlomo emphasized that Base1 aims to be "faster and cheaper for customers eventually than using the frontier models like Opus." In a press release, Base44 noted that owning its model would yield "direct control over compute and inference spend, expected to result in a structurally stronger margin profile over time."

The frontier threat#

Despite Base44's ambitions, risks remain. Major AI labs are encroaching on the vibe-coding space: Cursor and Grok's parent company xAI both , while Anthropic's Claude Code has become a serious competitor. These players have scale, resources, and feedback loops that smaller startups cannot easily replicate.