⌘K
Change language Switch ThemeSign In
Narrow Mode
Real-world UI Design Capabilities Comparison of Mainstream LLMs: Codex vs. Opus vs. Gemini ==========================================================================================
Real-world UI Design Capabilities Comparison of Mainstream LLMs: Codex vs. Opus vs. Gemini ==========================================================================================  ### Viking
@vikingmute
5.3 Codex 甚至是 5.4 现在的唯一短板就是UI设计能力比较差，页面真的丑，即使用了各种设计 skill 也还是不行，反观 4.6 Opus 用了impeccable.style 或者 taste skill，总是能一下非常惊艳我，所以我现在页面设计生成一般会交给 Opus 4.6 或者 Gemini 3.1
#### Viking
@vikingmute · 3d ago
感觉 Codex 已经是性价比最高的了，20美元每月，可以多端，有桌面应用，有 cli，有cloud agents，手机通过 ChatGPT 也能访问云端任务，有 code review，模型上来说 GPT-5.3-Codex / 5.4 也很能打，长任务效果也不错。介绍一个朋友用免费版，发现日常轻度使用也有不少额度。
最近几天已经没有用 Claude Code 了，封号，中转又贵。想用的时候就用我的 Cursor Legacy Plan 也可以很爽快的用 Opus 4.6。Show More
43
21
282
84.6K
Mar 14, 2026, 1:53 PM View on X
5 Replies
1 Retweets
34 Likes
3,863 Views V Viking @vikingmute
One Sentence Summary
Shares insights on the differing UI page design generation capabilities of GPT-5.3/5.4 Codex compared to Claude 4.6 Opus and Gemini 3.1.
Summary
Based on practical development experience, the user compares the UI design performance of several cutting-edge large language models. They point out that while GPT-5.3/5.4 Codex offers good value for money in coding tasks, it has a clear shortcoming in UI design aesthetics and generation quality. In contrast, Claude 4.6 Opus and Gemini 3.1 can produce more stunning pages when combined with specific design techniques. This provides practical reference for developers in choosing the most suitable AI model for different task scenarios.
AI Score
80
Influence Score 14
Published At Today
Language
Chinese
Tags
LLM Comparison
UI Design
GPT-5.3
Claude 4.6 Opus
Gemini 3.1 HomeArticlesPodcastsVideosTweets
Real-world UI Design Capabilities Comparison of Mainstrea... ===============