We pitted Base 44's new AI model against Anthropic's to build the same website. One was faster.

In a head-to-head comparison, Base44's newly launched AI model, Base 1, has demonstrated its potential to outperform Anthropic's Opus 4.8 in building websites. The San Francisco-based startup, led by founder and CEO Maor Shlomo, aims to distinguish itself from rivals such as Lovable, Replit, and Cursor by crafting unique design outputs that break away from the stereotypical 'AI-slop' aesthetic that has become prevalent in the industry. Shlomo explained that many websites generated with existing AI models share similar traits, often featuring bland color schemes and generic designs. He cited comments from design experts who have described these outputs as resembling an 'algorithmic Uniqlo or Ikea'—functional yet uninspired. With Base 1, Shlomo is optimistic about the future, stating that more advanced iterations will eventually yield designs that are distinctly different each time, although he acknowledges that achieving this goal will take effort. To test the capabilities of Base 1, I embarked on creating an e-commerce site for a fictional emotional support lamp that adjusts its color based on calendar events. After submitting my design prompt without any specific instructions, I opted for Base 1 as my AI model. Initial results showcased a deep blue background, a refreshing departure from the conventional light or beige designs typically associated with AI-generated websites, along with a bright yellow accent. Despite the distinct background, I noticed some trademark characteristics of AI-generated designs, including rounded corners and an abundance of emojis. To see how Base 1 measured up against Opus 4.8, I repeated the same prompt using the latter. Notably, Opus 4.8 took significantly longer to generate the site, with its first output revealing a familiar beige background, reinforcing the perception of generic design. Both models consumed the same amount of message credits, but when I requested an additional feature—color modes for exercise and eye breaks—Opus 4.8 used slightly more credits than Base 1 while taking longer to complete the task. Overall, while both models produced similar results, Base 1's faster output and efficiency in credit usage were evident. Looking ahead, Shlomo mentioned plans for 'reinforcement learning' to enhance Base 1's design capabilities, with hopes that future updates will yield even more innovative and unique website designs.