Base44, a young vibe-coding platform, has just unveiled its custom AI model to support app creation with natural language. This move comes as discussions around AI models’ suitability for specific use cases intensify, raising questions about long-term business defensibility. While Base44 hopes its model will outperform others in terms of latency and cost efficiency, other startups are following suit.

The founder, Maor Shlomo, believes that owning the model allows for more optimizations, but Jonathan Userovici from VC firm Headline points to data as a key component of defensibility. Base44’s first iteration, Base1, was trained on a dataset generated from tens of millions of real user interactions.

The bigger competition may come from AI labs like Anthropic and Claude Code, which are building more specialized models for app creation. Shlomo thinks specialization could give Base44 an edge over general-purpose models. Enterprise customers are also starting to demand cost optimization in AI usage.

Cost reduction is a key driver behind Base44’s decision, as ownership of the model gives them direct control over compute and inference spend. This could result in improved margins for the company, despite a delayed payoff. With plans to pass $100 million in annual recurring revenue, Base44 aims to cement its position as the only vertically integrated vibe-coding application.