Quantization at 1.58 bits

Ternary-weight models that retain most of full-precision quality.

Want the primary source?View original →

Related to

Implements

Related to (incoming)

⌥ PATH·GQuantization→Rquant-kit→·QuantBench→PQuantization at 1.58 bits

Related across the graph

Topics

Get the latest AI news, research, and insights delivered to your inbox.

Follow Angestrom

Global ingestion network

Continuous sync from primary AI sources — indexed, enriched, and queryable in real time.

arXivHugging FaceGitHubNewsFunding

Pipeline synced 24/7

ANGESTROM

The Intelligence Layer of Humanity. Everything AI. All in One Place.

Angestrom connects every piece of the AI ecosystem — data, models, research, companies, tools, and people.