AI Startup Race Intensifies as Base44 Launches Its Own AI Model

Base44, the AI-powered app-creation platform acquired by Wix for $80 million in June 2025 when it was just six months old with a team of eight, has launched its own proprietary large language model called Base1.

The model, now in production and actively serving users, was trained on a dataset generated from tens of millions of real user interactions on the platform, giving it unique training signals specific to Base44's use cases.

Founder Maor Shlomo framed the move as a strategic evolution: "Training and owning the model as part of [our] entire stack allows us a lot more optimizations on latency, cost, and efficiency".

"Having our own model means we can continue to improve it over time, making us less dependent on external vendors. This is not a trivial effort. It requires deep expertise, significant infrastructure, and access to the kind of rich, diverse data that most organizations simply do not have." – Maor Shlomo, Founder and CEO of Base44

The Defensibility Debate in AI

The decision comes amid an intensifying industry debate over whether businesses built on top of third-party frontier models can maintain a defensible long-term position.

Jonathan Userovici, a general partner at VC firm Headline, identifies three key ingredients for defensibility in AI startups: distribution, data, and tech stack.

By owning its model, Base44 gains control over all three, positioning itself as what Shlomo calls the "only vertically integrated vibe-coding.

The proprietary AI model trend is being driven by cost pressures and the realization that dependency on external providers creates margin erosion.

 Userovici noted that enterprise customers are increasingly questioning the return on investment from using the latest frontier models for every use case, driving demand for orchestration and optimization layers

Base44's ownership of its model gives it direct control over compute and inference spend, expected to result in a structurally stronger margin profile over time.

The move also reflects intensifying competition in the AI coding space. Swedish startup Lovable, which reached unicorn status in its Series A round, still relies on external LLMs and has reported $500 million in annual recurring revenue compared to Base44's $100 million. 

However, the bigger threat may come from frontier AI labs themselves. Anthropic's Claude Code, xAI's Grok, and Cursor are increasingly moving into vibe coding territory, giving them access to user data and feedback loops that can improve their own models.

Shlomo argues that specialization gives Base44 an edge, predicting that "models are progressing, but they'll stay very general in what they can do" .