Base44, the Vibe coding platform that Wix acquired for $80 million just a year ago, was only six months old, had a team of eight people, and had begun rolling out its own AI models to help users create apps using natural language.
The move comes as debate in the AI industry deepens about which frontier models are best for all use cases. A related question is whether a business built on someone else’s model is actually defensible in the long term. Bay Area-based Base44’s latest move speaks to both.
Although the custom LLM has just been released, Base44 expects it to eventually surpass the Frontier model. According to founder Maor Shlomo, “Training and owning models as part of the overall stack allows for much greater optimization for latency, cost, and efficiency.”
At first glance, this may be a way to get ahead of competitors like Lovable, a Swedish startup that reached unicorn status with a Series A round last summer and relies on external LLMs. But Shlomo expects others to train his models. “A player with enough scale and speed to at least have enough data.”
Data is one of three key pillars of defensibility for AI startups, along with distribution and technology stack, according to Jonathan Userovici, general partner at VC firm Headline, whose portfolio includes AI companies like Mistral AI but not Base44.
As a result, players with strong brands are now turning to data and infrastructure to boost their defenses, and Base44 fits this pattern. The company said Base1, the first iteration of LLM, was developed and trained on datasets generated from “tens of millions of real user interactions on the platform.”
This dataset will continue to grow with the company. But so will your competitors. The bigger competition may come not from Vibe coding startups, but from pioneering AI labs moving closer and closer to Base44’s home base. Cursor and Grok’s parent company, xAI, are now both part of SpaceX, and Claude Code has become a Vibe coding player in its own right.
This gives Anthropic and other base AI providers access to data and feedback loops they can use to improve their app creation models, but Shlomo believes specialization benefits Base44. “The models are evolving, but they will remain very general in terms of what they can do,” he predicted.
Userovici warned against underestimating pioneering models, citing the example of legal tech startup Harvey, which abandoned plans to educate its own model. He doesn’t expect applied AI companies to become cutting-edge labs all at once, but frames Base44’s move in a broader context where inference costs have become a meaningful part of the equation.
Userovici says these cost pressures have driven the changes now demanded by enterprise customers. “You don’t necessarily see a (return on investment) in using the latest model for every use case, so the entire infrastructure is being set up to do tuning and optimization to select the right model to ensure that costs don’t skyrocket while maintaining the same or similar performance for most use cases.”
Enterprise companies are still a small portion of the Vive Coding Platform’s audience, but they are making up a growing share of the platform’s revenue, and users of all sizes are starting to express concerns about the costs of using AI. Base44’s decision to develop its own LLM is due to a number of factors, but cost savings is likely to be one of the benefits.
“We want a model that is more in line with what we think is right, more optimized for what our users like in terms of the results we get, and ultimately faster and cheaper for our customers than using a front-line model like Opus,” Shlomo said.
For Base44 itself, the cost savings are less clear. In a press release, the company explained, “Model ownership will allow Base44 direct control over compute and inference spend, which is expected to result in a structurally stronger margin profile over time.”
Even if compensation is delayed, improved margins would be good news for Base44’s parent company, which recently announced it would lay off 20% of its workforce. By contrast, Base44 has been growing its headcount since the acquisition and announced a few months ago that it had surpassed $100 million in annual recurring revenue.
That’s still less than Lovable, which said it had hit $500 million in ARR earlier this month. However, Shlomo is confident that the “huge engineering effort” to develop Base1 will solidify Base44’s position as “the only vertically integrated Vibe coding application, or, in Userovici’s terms, player that simultaneously owns deployment, data, and infrastructure.”
If you purchase through links in our articles, we may receive a small commission. This does not affect our editorial independence.
