Maor Shlomo, Founder & CEO at Base44

A software company, in today’s age, is defined by engineering bets it’s willing to take and how far it’s willing to go to own what matters most.

Base44 has always been about taking on ambitious engineering projects. That has been true from day one, and it will continue to define how we build.

Today, we announced a major milestone: we trained our own model “Base 1”.

When Base44 was just getting started, one of the early choices was to build our own backend. It wasn’t the obvious path. The conventional wisdom was that a small team should never take on that much infrastructure. But the belief was simple: if Base44 was going to become the best way for people to build software, we needed to own the critical core of the product. That bet became one of our biggest advantages.

Today, we’re making a bigger bet of the same kind. We’re training our own model, purpose-built to create the next frontier of web applications.

This is the next step in a much bigger direction: owning the entire stack. Base44 is becoming a vertically integrated vibe coding platform, with its own database, its own backend, its own model, and the intelligence layer that connects all of it into one seamless building experience. We led with our own backend. Now we’re leading with our own model.

Why Base44 needs its own model

Frontier models are extraordinary, and we will continue to use them where they make sense. They are powerful because they are trained to help across an enormous range of tasks, from writing code to explaining concepts, summarizing documents, and reasoning through complex problems.

But that breadth is also the reason Base44 needs its own model. General models need to be good at everything. They need to understand many programming languages, many workflows, many domains, and many kinds of reasoning. 

Our focus is to help people build web apps inside an environment we understand deeply. Since day one, we’ve built feedback mechanisms directly into the agent, which means we can learn from millions of real building sessions across the platform. We can see what people asked for, what the agent built, what broke, what users changed, what they accepted or rejected, and most importantly, whether the app actually worked.

That last signal is incredibly rare. Most models have to optimize for a very wide range of use cases. We can optimize for something narrower and more specific.

That focus matters. A model trained specifically for building web applications, inside a harness designed specifically for apps, can become more powerful for this use case than a general-purpose model operating in a general-purpose environment. The harness is the layer around the model: the tools, context, workflows, and product environment that let the model act. Because Base44 controls that environment, we can train the model to perform better.

For our users, that means better applications, created faster and with more functionality. It means a model that is not just a better coder, but a better builder, one that understands what you are trying to create and helps you get there faster.

The missing piece in the Base44 stack

Base44 owns the entire product surface, the frontend and the backend running underneath it. The model was the missing piece, and is the intelligence at the center of the entire experience.

If the backend was the first major step toward owning the core of the product, meaning users no longer need to know what an API key is, what a database is, or how authentication works, then the model is the next one. It's what turns Base44 from a powerful building platform into a deeply integrated AI software creation system.

Owning it means we can shape the model around our users, our product, and the way people actually build on Base44. It also gives us more control over cost, latency, reliability, and quality, while still letting us use the best external models where they are the right fit.

As AI becomes a bigger part of how software is created, owning more of that intelligence becomes just as important as owning the infrastructure around it.

Teaching the model to think like a product partner

A great CTO tells you what to build. A great engineer writes the correct code. Most people who use Base44 aren’t CTOs or engineers. They have an idea and a goal, not a technical spec, and they don’t always know the right question to ask.

So we’re not only training a model to implement code correctly, we’re training it to make good product decisions, to understand what separates a great app from a mediocre one, to anticipate what you’ll need next, and to push back when there’s a better path.

The ambition is to build a different class of model: one that does great coding, but also takes product decisions with you. A model with taste. A model that can act less like a tool waiting for instructions and more like an expert helping you shape the right thing.

We study what makes the best apps on our platform succeed, and we use that to steer the model toward the choices that lead to working products. We want the model to be a trusted advisor, a technical partner that can help every builder make better decisions as they turn an idea into a real product.

The flywheel we’re building toward

This won’t happen overnight, and our first release isn’t trying to. The flywheel is simple and powerful. More building gives us more signal, more signal gives us a better model, and a better model brings more value for our users. Each checkpoint should be meaningfully better than the last, and we’ll share them with the world as we go.

We believe this can change what building software with AI feels like, not only inside Base44, but for anyone trying to turn an idea into a working application. And we can’t wait for you to build with it.

Our next big bet

We started Base44 so that anyone could build software, not just engineers or the people who can hire engineers. Training our own model is the most direct path we know to that goal.

Like the backend bet before it, we’re making it because we believe owning the core is worth it.

This is one of the most exciting engineering products I’ve worked on in my career, because it brings together everything Base44 was built for: ambitious engineering, deep product focus, and the belief that more people should be able to create software.

We’re building the model behind Base44 because we believe the future of software creation will belong to platforms that understand the entire building process, from the idea, to the product decisions, to the code, to the app that users actually experience.

Let’s build.