@kimmonismus· @kimmonismus · X·· 2026-08-28精选AI 评分71
AI 导读
GLM-5.3 已在 Hugging Face 正式上线，作者列出了在本地运行该模型的硬件门槛。FP8 需 10–12× H100 或 8× H200；4-bit/NVFP4 约 390–430GB，可用一台 512GB Mac Studio 或 4× DGX Spark；激进 2-bit 约 230–250GB，一台 256GB Mac Studio 或 2× DGX Spark，但有质量和上下文方面的权衡。首批 GGUF 和 NVFP4 量化已开始出现。
推荐理由
GLM-5.3 上线后给出了各量化档位的显存与硬件门槛，可供本地部署选型参考。
正文
GLM-5.3 is officially live on Hugging Face.
What should run it locally:
- FP8: 10–12× H100 or 8× H200
- 4-bit/NVFP4, roughly 390–430GB: one 512GB Mac Studio or 4× DGX Spark
- Aggressive 2-bit, roughly 230–250GB: one 256GB Mac Studio or 2× DGX Spark, with quality and context tradeoffs
The first GGUF and NVFP4 quants are already appearing.
Have fun with SOTA local AI
Quick reminder: in 4 hours you will be able to download and run sota AI. If you have enough compute ofc. What a time to be alive. https://t.co/T6bs1VnVG1在 X 查看被引用的帖子
来源：@kimmonismus · x.com