Base44 launches its own AI model as vibe coding platform seeks long-term defensibility

Base44, the vibe coding platform acquired by Wix for $80 million just one year ago, has begun rolling out its own proprietary AI model. The move signals a strategic shift for the young company, which was barely six months old with a team of eight at the time of acquisition. The decision comes amid a broader industry debate about whether businesses built on top of third-party frontier models can maintain a defensible long-term position.

A bet on vertical integration

Base44’s custom LLM, named Base1, is currently being introduced to its user base. The model was trained on a dataset generated from “tens of millions of real user interactions on the platform,” according to founder Maor Shlomo. The company believes that owning its model will allow for greater optimization in latency, cost, and efficiency compared to relying on external providers like OpenAI or Anthropic.

“Training and owning the model as part of our entire stack allows us a lot more optimizations,” Shlomo said. The goal is to eventually outperform frontier models for the specific use case of generating applications from natural language prompts.

The defensibility debate in AI

The move directly addresses a central question facing many AI startups: Can a company that depends on another firm’s foundational model build a sustainable business? Jonathan Userovici, a general partner at venture capital firm Headline, identifies three key ingredients for defensibility in AI: distribution, data, and tech stack. Base44’s decision to develop its own model allows it to own all three.

Userovici noted that inference costs have become a meaningful part of the equation for many companies. Enterprise customers, in particular, are increasingly questioning the return on investment from using the latest frontier models for every use case. This has driven demand for orchestration and optimization layers that can select the right model for the job, keeping costs manageable without sacrificing performance.

Competitive landscape and cost pressures

Base44’s vertical integration strategy could help it differentiate from competitors like Swedish startup Lovable, which reached unicorn status in its Series A round and still relies on external LLMs. However, the competitive threat also comes from frontier AI labs themselves. Companies like Anthropic (with Claude Code), xAI (with Grok), and Cursor are increasingly moving into the vibe coding space, giving them direct access to user data and feedback loops that can improve their own models.

Shlomo argues that specialization gives Base44 an edge. “Models are progressing, but they’ll stay very general in what they can do,” he predicted. Userovici, however, cautions against underestimating frontier models, pointing to the example of 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 toward cost optimization.

Financial implications for Wix

Base44’s model development comes at a time when its parent company, Wix, recently announced a 20% workforce reduction. In contrast, Base44 has been growing in headcount since the acquisition and reported surpassing $100 million in annual recurring revenue a few months ago. That figure still trails Lovable, which hit $500 million in ARR earlier this month.

The company stated that ownership of the model gives it direct control over compute and inference spend, which is expected to result in a structurally stronger margin profile over time. While the payoff may be delayed, improved margins would be welcome news for Wix.

Conclusion

Base44’s launch of Base1 represents a significant bet on vertical integration in the AI application layer. By owning its distribution, data, and infrastructure, the company aims to build a defensible position in the increasingly crowded vibe coding market. Whether this strategy will pay off depends on its ability to keep pace with rapid improvements in frontier models while maintaining cost advantages for its users.

FAQs

Q1: What is Base44’s new AI model called?
Base44’s custom LLM is named Base1. It was trained on data from tens of millions of real user interactions on the platform.

Q2: Why did Base44 decide to build its own model?
The company aims to reduce costs, improve latency, and increase control over its technology stack. This is part of a broader strategy to build a defensible business that doesn’t rely solely on third-party frontier models.

Q3: How does this compare to competitors like Lovable?
Lovable, a Swedish vibe coding startup that reached unicorn status, still relies on external LLMs. Base44’s vertical integration approach could provide cost and performance advantages, but Lovable has a much larger revenue base, with $500 million in ARR compared to Base44’s $100 million.