Base44 founder Maor Shlomo has started rolling out Base 1, a proprietary AI model for the Wix-owned app-building platform, a move that turns his original no-code bet into a vertical-integration bet.
The launch, reported by TechCrunch on June 29, 2026, is not simply a model announcement. It is Shlomo's answer to the question hanging over every fast-growing AI application company: whether a startup that depends on frontier models for its core workflow can keep enough margin, data advantage and product control to matter once the model providers move closer to the user.
Base44 says Base 1 has started rolling out in production and is serving users on the platform. In Base44's announcement, the company described the model as a general-purpose AI agent for conversation and coding, including multi-turn interactions, tool use, runtime management and backend operations. Base44 also says the model was developed from production traffic and "tens of millions of real user interactions" on its platform, a company-supplied figure that has not been independently benchmarked against the frontier models Base44 still uses where it makes sense.
That caveat matters. Base44 has not disclosed Base 1's architecture, size, training cost, benchmark performance, or the share of user traffic now routed to the new model. What it has disclosed is the direction of travel: less dependence on outside model vendors, more control of inference costs, and a tighter loop between the apps users ask for, the apps Base44 generates, and the model that learns from those outcomes.
The founder's second software-building company
Shlomo is not a first-time founder discovering infrastructure late. Before Base44, he co-founded Explorium in 2017 with Or Tamir and Omer Har and served as CEO. In a CTech interview, Shlomo said his first taste of product-building came in Israeli Military Intelligence, where he taught himself programming and built an internal analysis tool that he said reached 12,000 monthly users.
The origin of Base44 was more personal and more instructive. After reserve duty in Unit 8200 following the October 7, 2023 attacks, Shlomo stepped back from day-to-day work at Explorium while remaining a shareholder and board member, then traveled with his partner, tattoo artist Yuval Dahan. CTech reported that he tried to build her a website and customer-management system and found the process needlessly difficult, even as the CEO of a tech company. His conclusion was not that LLMs could not write code. It was that they lacked the right environment to turn intent into working software.
That thesis became Base44: describe an app in natural language and let the platform generate the front end, backend, database, authentication, deployment and related workflows. Base44's own site says users can build apps by describing an idea in plain language, then iterate by chatting with the AI; it also says Base44 automatically sets up user logins, authentication, data storage and role-based permissions behind the scenes.
The Base 1 launch extends that same founder instinct. In a blog post about the model, Shlomo wrote that Base44 first made the non-obvious choice to build its own backend, then framed Base 1 as the next step toward "owning the entire stack." His argument is that a model trained inside Base44's harness can learn not just what code compiles, but what product choices led to working apps on Base44.
Wix bought the company. It also bought the margin problem.
Wix announced the Base44 acquisition on June 18, 2025, a year before this model rollout, so the deal should not be treated as fresh news. Wix's press release described Base44 as an AI-powered platform for building custom software and apps with natural language.
The economics behind the deal are clearer in Wix's annual filing than in the acquisition headline. Wix's 2025 annual report says it acquired 100% of Base44 Inc. on June 13, 2025 for total purchase consideration of $92.158 million, including $18.058 million in cash and $74.1 million in acquisition-date fair-value contingent consideration. The filing also warns that Base44 operates in a short-history, early-stage vibe-coding market where AI infrastructure, data acquisition and specialized talent costs are substantial, and where competition from dominant LLM developers could limit monetization.
That is the strategic subtext of Base 1. For Wix, Base44 is both growth story and cost center. Wix said in its first-quarter 2026 results that Base44 had achieved about $150 million of ARR as of May 2026. Base 1 applies a similar logic to the app-builder surface: a domain-specific model may not need to be the best model in the world if it can be cheaper, faster and better-tuned for one workflow.
Base44's pricing page shows why inference discipline matters. The product has a free tier and paid annual plans listed at $16, $40, $80 and $160 per month, with message credits and integration credits gating usage. Its FAQ says integration credits are consumed when an app calls integrations such as LLMs, file uploads, image understanding, image generation, email sending or SMS. In other words, every successful builder potentially creates more downstream compute exposure.
The defensibility argument is real, but not settled
Base44's strongest claim is not that it has replaced OpenAI, Anthropic or other frontier labs. It is that it has a workflow they do not fully own: users asking for apps, Base44 generating those apps, users accepting, rejecting or revising the outputs, and the platform seeing whether the resulting software works.
That is valuable data if the loop is large enough and clean enough. TechCrunch quoted Headline general partner Jonathan Userovici, whose firm is not a Base44 investor, saying data is one of three core ingredients for AI defensibility, alongside distribution and tech stack. Base44 is trying to assemble all three under Wix: distribution through a fast-growing app-builder, data from production sessions, and infrastructure that now includes a proprietary model.
The counterpoint is just as important. Frontier model companies are not standing still, and coding is one of the highest-value use cases in AI. Anthropic has Claude Code. Cursor, Replit, Lovable, Bolt.new and Vercel's v0 all sit somewhere along the spectrum from coding assistant to app generator. Some depend heavily on third-party models; others are wrapping models inside tighter development environments. The more the frontier labs package their models into end-to-end coding workflows, the more Base44's differentiation has to come from the product layer and accumulated app-building data, not from model ownership alone.
There is also an operational trust issue that comes with becoming the place where non-technical users build software. In July 2025, Wiz disclosed a critical Base44 vulnerability that could have allowed unauthorized registration into private apps. Wiz said Wix fixed the issue within 24 hours and that Wix found no evidence of malicious exploitation. The episode did not stop Base44's growth, but it illustrates the burden that comes with Shlomo's premise: if Base44 abstracts away engineering for users, Base44 inherits more of the engineering risk.
Shlomo's bet is that this trade is worth it. Base44 began as a way to remove programming and setup barriers for people with ideas but no engineering team. Base 1 makes the same bet at a deeper layer: that the app builder should not merely rent intelligence from the frontier labs, but teach its own model from the building sessions happening inside its walls.
That is an expensive route for a young product, even under Wix. It is also the route that gives Base44 the clearest answer to the defensibility question. If Shlomo is right, the moat is not just the model. It is the loop between the builder, the app, the runtime and the model learning from all three.