想找能照着截图做网页的模型?直接看 Code Arena 这份榜单,Opus 5 (Max) 第一,GPT-5.6 和 Grok 的分数也都列出来了。
Code Arena 公布 Image-to-WebDev 最新评测分数。Opus 5 (Max) 以 1669 分排名第一。GPT-5.6 的 Sol、Terra、Luna 三个版本分别拿到 1581、1531、1497 分,排在第 4、13、21 位。Grok-4.5 以 1578 分排第 5,Kimi K3 (Max) 以 1571 分排第 6,Muse Spark 1.1 以 1550 分排第 8。该评测考察模型根据截图生成网站,以及多步推理和工具调用的智能体编码流程。
Update: Scores are in for Image-to-WebDev in Code Arena for Opus 5 (Max), GPT-5.6 Sol, Luna and Terr...
Update: Scores are in for Image-to-WebDev in Code Arena for Opus 5 (Max), GPT-5.6 Sol, Luna and Terra, Grok-4.5, Kimi K3 (Max), and Muse Spark 1.1. See how they stack up. - #1 Opus 5 (Max) with 1,669 pts - #4 GPT-5.6 Sol (xHigh) with 1,581 - #5 Grok-4.5 with 1,578 pts - #6 Kimi K3 (Max) with 1,571 - #8 Muse Spark 1.1 with 1,550 pts - #13 GPT-5.6 Terra with 1,531 pts - #21 GPT-5.6 Luna with 1,497 pts Image-to-WebDev in Code Arena, allows you to view overall rankings across AI models on their ability to generate websites from images and screenshots, alongside agentic coding workflows that involve multi-step reasoning and tool use. Learn more below. 💬 3 🔄 1 ❤️ 11 👀 1486 📊 4 ⚡
- AI Will08:38原文