728 x 90

We tested the first Base44 LLM model to create a website

We tested the first Base44 LLM model to create a website

Base44 launched its own AI model to outperform competitors like Lovable, Replit, and Cursor. Maor Shlomo, founder and CEO of Base44, told Business Insider that part of the reason the San Francisco-based vibration coding startup formed its own LLM was to reduce the veneer of AI that has come to define vibration-coded products. Design gurus

Base44 launched its own AI model to outperform competitors like Lovable, Replit, and Cursor.

Maor Shlomo, founder and CEO of Base44, told Business Insider that part of the reason the San Francisco-based vibration coding startup formed its own LLM was to reduce the veneer of AI that has come to define vibration-coded products.

Design gurus have said that websites built with cutting-edge models like Anthropic’s Opus 4.8, OpenAI’s GPT-5.5, or Gemini 3 often share similar interfaces and design elements, such as rounded corners, beige color palettes, and excessive use of emojis. Base44’s competitors, such as Lovable and Replit, use these underlying models.

For example, Paul Bakaus, CEO of AI design startup Impeccable, said in a June interview with Andreessen Horowitz that the designs of these sites look like an “algorithmic Uniqlo or Ikea”: functional but basic.

Shlomo said Base 1, in its more advanced versions in the future, “will create something that looks exceptionally different” every time, but “it’s going to take some effort, so we’re not there yet.”

He also said the model would produce faster results and cost users fewer credits than using frontier models. Base44 is a subsidiary of website building company Wix.

To see how the model fared, I pitted Base 1 against Anthropic’s Opus 4.8 to code the same website.

Create a website for a fictional mood lamp


I could select Base 1 as my functional AI model.

I selected Base 1 as my functional AI model while building my website.

Screenshot/Business Insider



I decided to create an e-commerce website for a fictional product: an emotional support lamp that can read Google calendars and change its tone depending on the current event. I asked it to light up yellow for focus mode, blue for downtime, pink for meetings, and red for looming deadlines.

I gave Base44 my message, carefully avoiding any instructions on website design, curious to see what design it would come up with on its own.

Before pressing Enter, I selected Base 1 as my AI model. Free users do not have this option; only users with the $40 per month “Builder” subscription or higher can switch between AI models.

After a few minutes of anticipation, Base44 introduced me to the “Lumos” site.

Base 1 vs Opus 4.8


A screenshot of a vibrate coded website created by Base44.

I was surprised by the deep color palette, one I hadn’t spontaneously seen on a vibe-coded website.

Screenshot/Business Insider



From the beginning, I was surprised by the design. The website had a deep blue background, already setting it apart from the light or beige interfaces I had seen on most vibe-coded websites. It was complemented by a bright yellow accent.

The rest of the website was basic, with some telltale signs of being vibration-coded. For example, elements had rounded corners and a lot of emojis were used.


Base44's Base 1 generated a website with lots of emojis.

Base44’s Base 1 model generated a website with lots of emojis.

Screenshot/Base44



As Shlomo said, Base 1 has a long way to go before its designs can be truly unique.

For comparison, I sent Base44 the same message again, but this time I chose Anthropic’s Opus 4.8 as the working model.

The first thing that caught my attention was that Opus 4.8 took longer to generate the site than Base 1, which seemed twice as long. When it finally loaded, I had my first “aha!” moment – hello, beige background.


A screenshot of the Base44 encoded website.

The color palette of the website coded in Opus 4.8 was more than what I was used to on vibrate coded websites.

Screenshot/Business Insider



Aside from the color palette, everything else was similar to the website coded in Base 1. While Base 1 was faster, they both used the same amount of credits, 1.2 message credits each of my 250 monthly credits.

I then tried to modify both websites to add an additional feature. I asked both models to add a few more color modes: green for exercise and teal for resting your eyes. To complete this task, Opus 4.8 consumed 1.4 credits of messages, while Base 1 did the job in 1.2 credits and made the change faster.

While the difference was small, Opus was proving to use credits faster for what was effectively the same adjustment.

All things considered, I didn’t see a major difference in how the two models performed, other than the design generated by Opus seemed a bit more generic.

Shlomo said the Base44 team will conduct “reinforcement learning” on Base 1, which involves prompting the model to keep generating designs that look new and unique.

We’ll check back when the next version is released.