Our Smallest, Most Efficient Model: Muse Ember

Re:Our Smallest, Most Efficient Model: Muse Ember by loeg on Monday November 09, 2026 @04:11PM (#50432210)

Spark Open was just October. Give it time. https://research.meta.ai/blog/muse-spark-open

[ Parent | on HN ]

Re:Our Smallest, Most Efficient Model: Muse Ember by kristianp on Monday November 09, 2026 @05:44PM (#50433363)

Which was 816B parameters.

[ Parent | on HN ]

Re:Our Smallest, Most Efficient Model: Muse Ember by anon373839 on Monday November 09, 2026 @06:14PM (#50433715)

It was also nerfed to the point of no longer being in a comparable intelligence class to the original.

[ Parent | on HN ]

Re:Our Smallest, Most Efficient Model: Muse Ember by Schlagbohrer on Tuesday November 10, 2026 @03:46AM (#50437659)

How do you know they nerfed it? That must have happened before the weights were opened, right?

[ Parent | on HN ]

Re:Our Smallest, Most Efficient Model: Muse Ember by rsfern on Tuesday November 10, 2026 @04:28AM (#50437936)

The open version model card says it had extra safety post-training because meta can’t prefilter prompts with a classifier model if you’re running it on your own hardware. They don’t want to get sued I guess if it comes out that someone used their model irresponsibly or something

[ Parent | on HN ]

Re:Our Smallest, Most Efficient Model: Muse Ember by loeg on Monday November 09, 2026 @07:04PM (#50434247)

We don't know how many parameters the closed models have, so what are you comparing this number against?

[ Parent | on HN ]