Base44 launches custom AI model to strengthen defensibility
Original: Vibe coding platform Base44 launches own model as AI startups seek defensibility
Why This Matters
Illustrates AI startup trend of building proprietary models for defensibility amid rising inference costs and frontier model competition.
Base44, a vibe coding platform acquired by Wix for $80 million in 2025, has begun rolling out its own AI model to help users create apps using natural language, addressing industry concerns about long-term defensibility of businesses built on third-party models.
Base44, a Bay Area-based vibe coding platform, has started deploying its custom large language model called Base1 to support its user base in natural language app creation. The move reflects broader industry discussion about whether frontier models serve all use cases optimally and whether companies built atop external models can maintain defensibility. Founder Maor Shlomo stated that owning the model as part of the company's entire stack enables significant optimizations in latency, cost, and efficiency. Base1 was developed and trained on a dataset derived from tens of millions of real user interactions on the platform, which will continue to grow with the company. The move positions Base44 against competitors like Swedish startup Lovable, which reached unicorn status in its Series A round and relies on external LLMs. Jonathan Userovici, a general partner at VC firm Headline, identifies three key defensibility ingredients for AI startups: data, distribution, and tech stack. However, Base44 faces competition not only from vibe-coding rivals but also from frontier AI labs—Cursor and xAI (SpaceX subsidiary) now share ownership, and Claude Code has become a vibe coding player. Shlomo believes specialization provides advantage despite ongoing model improvements, while Userovici cautions against underestimating frontier models' capabilities, noting that legal tech startup Harvey abandoned its own model training plans.