Base44, the vibe coding platform acquired by Wix for $80 million just one year ago—when it was barely six months old and had a team of eight—has begun rolling out its own AI model to help users build applications using natural language. This move comes as debate intensifies in AI circles over whether frontier models are suitable for all use cases, and whether businesses built on third-party models can achieve long-term defensibility. Base44’s decision speaks to both concerns.
The Bay Area-based company’s custom large language model (LLM), named Base1, is currently rolling out. Founder Maor Shlomo expects it to eventually outperform existing frontier models: “Training and owning the model as part of our entire stack allows us a lot more optimizations on latency, cost, and efficiency.”
At first glance, this could help Base44 stay ahead of rivals like Swedish startup Lovable, which reached unicorn status last summer and now reports $500 million in annualized revenue, relying on external LLMs. However, Shlomo anticipates that other players will train their own models—"at least those with enough scale and velocity to have sufficient data."
According to Jonathan Userovici, a general partner at VC firm Headline (whose portfolio includes Mistral AI but not Base44), data is one of three key pillars of defensibility for AI startups, alongside distribution and tech stack. The implication: strong brands are increasingly leaning into their data and infrastructure to bolster defensibility. Base44 fits this pattern, with Base1 trained on a dataset generated from tens of millions of real user interactions on its platform.
As Base44’s dataset grows, so will those of its competitors. The bigger challenge may come not from vibe-coding startups but from frontier AI labs encroaching on Base44’s territory. Notably, Cursor and xAI’s Grok now both belong to SpaceX following a blockbuster acquisition in June 2026, and Claude Code has emerged as a vibe coding player in its own right. This gives Anthropic and other foundational AI providers access to data and feedback loops for improving app creation models. Yet Shlomo argues that specialization gives Base44 an edge: “Models are progressing, but they’ll stay very general in what they can do.”
Userovici warns against underestimating frontier models, citing legal tech startup Harvey, which abandoned plans to train its own model. He does not expect applied AI companies to become frontier labs en masse but sees Base44’s move as part of a broader trend where inference plays a growing role in competitive differentiation.
via TechCrunch AI