宝玉表示，Fable 5.5 的视频能力有所提升，但对普通用户来说收益有限

发生了什么

Fable 5.5 的视频能力进一步提升了；它确实很强，仅靠代码就达到了这种水平。 但对普通人来说，这项能力的提升有限，我更期待它在其他方面有所改进。 引用 @chetaslua 的帖子：Fable 5.5，视频和图像模型的时代结束了 🤯 这纯粹是代码，一次生成了这段视频，太他妈惊艳了，怎么会比 AI 视频模型还漂亮。你可以通过两种方式亲自验证：在关闭网页搜索和记忆的情况下，提问：“@karpathy 2026 年去了哪里？”只有新版

接下来读什么

• 今日还有哪些重要变化通过五项核心价值观解读 OpenAI 更新后的 Model Spec

2 篇报道

这件事到目前为止

Fable 5.5, an unreleased model that an author appears to have encountered in gray-scale testing, generated a video from pure code that impressed commenters, though Baoyu said on X that the gain is limited for ordinary users and that he would rather see other capabilities improve. The test used the prompt "Where did @karpathy go in 2026?" with web search and memory turned off. Chetaslua, whose post Baoyu quoted, called the result "pure code" that was "one-shotted" and said it looked more beautiful than AI video models. That is a promotional claim about the output, not a confirmed measure of quality. Baoyu's assessment is narrower: he said Fable 5.5's video capability has improved further and is genuinely strong for being achieved with pure code, but he framed the benefit to ordinary users as limited.