Vibe Coding's Base44 Launches Own AI Model for Defensibility

Base44, the innovative "vibe coding" platform acquired by Wix for $80 million just a year ago—when it was a nascent company with only eight employees and six months of operation—has initiated the rollout of its proprietary AI model. This new model is designed to empower users in developing applications through natural language interactions.

This strategic move by the Bay Area-based company emerges amidst intensified discussions within AI circles concerning the universal applicability of frontier models and the long-term defensibility of businesses built entirely upon third-party AI solutions. Base44's decision directly addresses both of these critical industry questions.

While its custom Large Language Model (LLM) is in its initial deployment phase, Base44 harbors ambitions for it to eventually surpass the performance of existing frontier models. According to founder Maor Shlomo, "training and owning the model as part of [our] entire stack allows us a lot more optimizations on latency, cost, and efficiency."

Superficially, this could be perceived as a tactic to gain an edge over rivals like the Swedish startup Lovable, which achieved unicorn valuation in its Series A round last summer and currently relies on external LLMs. However, Shlomo anticipates that other significant players, particularly those with sufficient scale and data velocity, will also develop their own models.

Jonathan Userovici, a general partner at VC firm Headline—whose investment portfolio includes prominent AI companies such as Mistral AI, though not Base44—identifies data as one of three fundamental pillars for AI startup defensibility, alongside robust distribution channels and a strong tech stack.

The implication is that companies with established brands are increasingly leveraging their unique data and infrastructure to fortify their market position, a pattern Base44 exemplifies. The company states that its inaugural LLM iteration, named Base1, was meticulously developed and trained using a vast dataset derived from "tens of millions of real user interactions on the platform."

While Base44's proprietary dataset will continue to expand, so too will those of its competitors. The more formidable challenge might not originate from other vibe-coding startups, but rather from frontier AI labs encroaching on Base44's core domain. Notably, Cursor and Grok's parent company xAI are now both affiliated with SpaceX, and Claude Code has independently emerged as a formidable player in the vibe coding landscape.

This evolution provides foundational AI providers like Anthropic with access to valuable data and feedback loops for refining app creation models. Yet, Shlomo believes Base44's specialized focus offers a significant advantage, predicting that while "models are progressing, but they’ll stay very general in what they can do."

Userovici, however, cautioned against underestimating the capabilities of frontier models, referencing the legal tech startup Harvey, which ultimately abandoned its plans to train a bespoke model. While he doesn't foresee a mass migration of applied AI companies into becoming frontier labs, he contextualizes Base44's move within a broader trend where inference costs have become a critical factor.

This escalating cost pressure, Userovici explains, has precipitated a shift in demand from enterprise clients. He notes, "They don’t necessarily see a [return on investment] when using the latest models for all use cases, so an entire infrastructure is being set up to do orchestration and optimization to select the right models for them so that costs don’t skyrocket while maintaining the same or similar performance across the majority of use cases."

Although enterprise companies currently represent a minority among vibe coding platform users, they account for a growing proportion of platform revenue. Furthermore, users across all segments are voicing concerns regarding the escalating costs associated with AI usage. Base44's decision to develop its own LLM was multi-faceted, with cost reduction undoubtedly being a key anticipated benefit.

Shlomo articulated the company's objectives, stating, "We want to get a model that is going to be more aligned to what we think is the right thing, is going to be more optimized to what we see users like in terms of the results we’re getting, and is going to be faster and cheaper for customers eventually than using the frontier models like Opus."

For Base44 itself, the immediate impact on cost reduction is less straightforward. In a press release, the company elucidated that "ownership of the model gives Base44 direct control over compute and inference spend, expected to result in a structurally stronger margin profile over time."

Even with a delayed return, enhanced margins would be welcome news for Base44's parent company, Wix, which recently announced a 20% workforce reduction. In stark contrast, Base44 has experienced consistent headcount growth since its acquisition and, a few months prior, reported surpassing $100 million in annual recurring revenue (ARR).

This ARR figure is still below Lovable's reported $500 million earlier this month. Nevertheless, Shlomo is confident that the "huge engineering effort" invested in developing Base1 will solidify Base44's position as the "only vertically integrated vibe-coding application"—a distinction that, in Userovici's framework, signifies a player possessing simultaneous ownership of its distribution, data, and infrastructure.