MiniMax (official)· @MiniMax_AI · X·· 2026-08-28AI 评分42
AI 导读
MiniMax 发文祝贺 fal 团队基于 MiniMax H3 推出 H3 Max。fal 对 H3 进行后训练以提升提示词遵循与视觉质量，并协同设计推理栈，在提速同时持续测试质量，仅保留不损失后训练收益的优化。
正文 · 原文
A great deep dive from the @fal team on how they built H3 Max, combining post-training with a co-designed inference stack to improve prompt adherence, visual quality, and speed.
We're proud to see MiniMax H3 serve as the foundation for this work. Congratulations to the team, and thank you for pushing H3 forward! 👏
Yesterday, we launched H3 Max. We post-trained MiniMax H3 for stronger prompt adherence and visual quality, while co-designing the inference stack alongside it. As we pushed speed, we continuously tested for quality, only keeping optimizations that preserved the gains from post-training. More on how we built H3 Max: https://blog.fal.ai/introducing-h3-max-by-fal/在 X 查看被引用的帖子
来源：MiniMax (official) · x.com