Comprehensive Review of AI Performance in UI/UX Design and Coding: Insights from a Global Survey
In recent months, I embarked on an extensive research project involving over 6,000 participants worldwide to evaluate how various Artificial Intelligence models perform in the realms of user interface (UI)/user experience (UX) design and programming. The data gathered is entirely open-source, with all model outputs and user interactions freely accessible. This initiative is purely academic, and I do not derive any financial gain from it.
To facilitate objective comparisons, I developed a crowdsourced benchmarking platform—Design Arena—where users can generate a variety of digital assets such as websites, games, 3D models, and data visualizations. They can also compare outputs across different AI models to determine which performs best in specific contexts.
Since launching, nearly 4,000 votes have been cast by roughly 5,000 active users, offering valuable insights into the strengths and weaknesses of leading AI solutions. Here are the key findings:
Top Performing Models for Coding and Design
The standout performers in both programming and design tasks are the Claude and DeepSeek models. Among these, Claude Opus emerged as the most favored by users, earning high marks for its interface and coding capabilities. The top-tier models also include DeepSeek v0—particularly praised for website development—and a surprising dark horse, Grok. However, it’s worth noting that DeepSeek models tend to be slower, which may influence your choice depending on your project needs.
The Underrated Contender: Grok 3
Grok 3 deserves special mention as an underrated yet highly capable model. Despite limited online visibility—possibly due to associations with Elon Musk—the model consistently ranks within the top five and offers notably faster performance compared to its competitors, making it a strong candidate for those prioritizing speed.
Mixed Results: Gemini 2.5-Pro and Other Models
Gemini 2.5-Pro has elicited mixed responses. Some users have expressed concerns about its lower ranking, questioning its reliability. While it excels at certain UI/UX tasks, it occasionally produces poorly designed applications. Nonetheless, it demonstrates proficiency in coding business logic, making it situationally useful.
AI Giants: OpenAI’s GPT and Meta’s Llama
OpenAI’s GPT models perform adequately, placing around the middle of the leaderboard. Conversely, Meta’s Llama models lag significantly behind, underscoring the competitive gap
