Vibe-coding platform Base44 is rolling out its own trained model, Base1, a departure from competitors that rely on frontier LLMs from Anthropic or Google. The Tel-Aviv based startup, which helps users create apps through simple prompting, aims to address concerns about using external models. Base44 was founded by Maor Shlomo, who said “training and owning the model as part of [our] entire stack allows us a lot more optimizations on latency, cost, and efficiency.” Launched in February 2025, Base44 came about because Shlomo was trying to help his girlfriend set up a website. Four months after creating Base44, Shlomo sold his company for an $80 million cash deal. The company recently passed $150 million in annual recurring revenue. Launching Base1 fills the only “missing piece” in Base44’s offerings, the company said. Competitors Sweden-based Lovable uses Claude and California-based Replit uses a combination of Claude and Gemini. Base44 believes Base1 will help it compete with Lovable, which leads at $500 million ARR as of this month. The in-house model will also lead to cost reductions by giving “direct control over compute and inference spend, expected to result in a structurally stronger margin profile over time,” Shlomo said.