返回
原文
Rohan Paul@rohanpaul_ai
34AI 编辑部评分，满分 100
2026-08-05 10:48· 51分钟前
AI 导读

阿里 Qwen3.8-Max 在 Arena 图像转网页开发（Image-to-WebDev）排行榜位列第二，得 1631 分，仅落后 Claude Opus 5 (Max) 39 分。该基准测试要求模型根据图像或截图生成网站，考验视觉理解、前端编码与多步工具调用。Qwen3.8-Max 为 2.4T 参数 MoE 模型，支持图像输入与 1M token 上下文窗口。

Alibaba's Qwen3.8-Max entered Arena's Image-to-WebDev leaderboard at #2.

behind Claude Opus 5 （Max） but ahead of Fable 5.

The benchmark gives models an image or screenshot and asks them to generate a website， combining visual interpretation， frontend coding， and multi-step tool use.

Qwen3.8-Max itself is a 2.4T-parameter MoE model supporting image input and a 1M-token context window.

Arena.aiExciting news: Qwen3.8-Max by @Alibaba_Qwen is #2 in Image-to-WebDev Arena! With 1,631 pts, it's trailing only Claude Opus 5 (Max) by 39 pts! Congrats again to ...
多模态编码评测/基准

来源：Rohan Paul · x.com

导出 Markdown
返回
34在 X 查看原推
导出 Markdown
Rohan Paul · @rohanpaul_ai · X·2026-08-05 10:48·51分钟前
AI 导读

阿里 Qwen3.8-Max 在 Arena 图像转网页开发（Image-to-WebDev）排行榜位列第二，得 1631 分，仅落后 Claude Opus 5 (Max) 39 分。该基准测试要求模型根据图像或截图生成网站，考验视觉理解、前端编码与多步工具调用。Qwen3.8-Max 为 2.4T 参数 MoE 模型，支持图像输入与 1M token 上下文窗口。

Alibaba's Qwen3.8-Max entered Arena's Image-to-WebDev leaderboard at #2.

behind Claude Opus 5 （Max） but ahead of Fable 5.

The benchmark gives models an image or screenshot and asks them to generate a website， combining visual interpretation， frontend coding， and multi-step tool use.

Qwen3.8-Max itself is a 2.4T-parameter MoE model supporting image input and a 1M-token context window.

Arena.aiExciting news: Qwen3.8-Max by @Alibaba_Qwen is #2 in Image-to-WebDev Arena! With 1,631 pts, it's trailing only Claude Opus 5 (Max) by 39 pts! Congrats again to ...

来源：Rohan Paul· x.com

多模态编码评测/基准