Base44, the vibe coding platform acquired by Wix for $80 million last year, has begun rolling out its own AI model to help users build apps using natural language.
The company, founded by Maor Shlomo, started with a team of eight and was barely six months old at the time of the deal. This shift addresses questions about whether businesses built on external models can survive long-term and whether frontier models fit every use case.
Why build a model?
While the custom LLM is only just available, Base44 hopes it will eventually outperform external options. Shlomo says training and owning the model allows for better optimisation of latency, cost, and efficiency.
This could help Base44 stay ahead of rivals like Lovable, a Swedish startup that reached unicorn status in its Series A round last summer and relies on external LLMs. Shlomo expects other players with enough scale and data to follow suit.
Jonathan Userovici, a general partner at VC firm Headline, notes that data is one of three key ingredients of defensibility for AI startups, alongside distribution and tech stack. The firm’s portfolio includes Mistral AI, but not Base44.
Base44 says the first iteration of its LLM, Base1, was trained on a dataset generated from tens of millions of real user interactions on the platform. This dataset will grow as the company expands.
Who are the real rivals?
The bigger competition may not be other vibe-coding startups. Cursor and xAI, the parent company of Grok, both belong to SpaceX. Claude Code has also become a vibe coding player.
This gives Anthropic and other foundational providers access to data and feedback loops to improve models for app creation. Shlomo thinks specialisation gives Base44 an advantage. He predicts models will progress but remain very general in what they can do.
Userovici cautions against underestimating frontier models. He cites Harvey, a legal tech startup that abandoned plans to train its own model. He does not expect applied AI companies to become frontier labs en masse.
He frames Base44’s move in a broader context where inference costs have become a meaningful part of the equation. Enterprise customers now demand orchestration and optimisation to select the right models. This prevents costs from skyrocketing while maintaining performance.
Enterprise companies remain a minority among vibe coding platform users, but they represent a growing share of revenue. Users of all sizes are expressing concerns over the cost of using AI.
Shlomo says the new model will be more aligned with user preferences and faster and cheaper than using frontier models like Opus. Base44 claims ownership gives it direct control over compute and inference spend, expected to result in a structurally stronger margin profile over time.
Even with a delayed payoff, improved margins would help Wix. The parent company recently announced it would lay off 20% of its workforce. In contrast, Base44 has been growing in headcount since the acquisition. The company passed $100 million in annual recurring revenue a few months ago.
Lovable reported hitting $500 million in ARR earlier this month. Shlomo bets the huge engineering effort to develop Base1 will cement Base44’s positioning as the only vertically integrated vibe-coding application. This means owning its distribution, data, and infrastructure all at once.
What it means
Base44 is betting that owning the model will reduce costs and improve margins over time. This approach aligns with a trend where companies with strong brands lean into their data and infrastructure to increase defensibility.
