$150M ARR Base44 rolls out its own AI model

Base44 is rolling out a custom AI model to cut costs and strengthen its vibe-coding platform.

Base44, the Wix-owned vibe-coding platform, has begun rolling out its own AI model as of June 29, 2026. The company says the move should improve latency, cost, and efficiency, while giving it more control over the stack behind app creation with natural language.

What changed

Base44’s first in-house LLM, called Base1, is trained on data generated from tens of millions of real user interactions on the platform. Founder Maor Shlomo says owning the model lets the company optimize for latency, cost, and efficiency rather than relying entirely on frontier models.

The company is still rolling out the model, but it expects the system to get better as its dataset grows. Base44 says the goal is not just lower spend, but a model tuned to the kinds of app-building outputs its users want.

• Base44 was acquired by Wix for $80 million one year ago.
• The startup had been around for about six months at the time of acquisition.
• It had a team of eight when Wix bought it.
• Base44 said in May it had passed $150 million in annual recurring revenue.
• That was up from $100 million just two months earlier.

Why it matters

The move reflects a wider AI startup push to build defensibility around data, distribution, and infrastructure. For vibe-coding tools, that matters because the biggest cost pressure now comes from inference, and customers are increasingly sensitive to whether the latest model is worth the price.

Base44 is also positioning itself against rivals that still depend on outside models, including Lovable. At the same time, frontier labs such as Anthropic and xAI are moving closer to app-building use cases, which raises the bar for specialized products that want to keep their edge.

The business case is simple: if Base44 can run a model that is faster and cheaper for its core use cases, it may protect margins as revenue grows. Wix has its own reasons to care, especially after announcing plans to cut 20% of its workforce.

The bigger question is whether more applied AI companies will follow Base44 and train their own models, or whether frontier providers will keep absorbing the best use cases before startups can build lasting separation.