Tag

#model efficiency

2 articles

NVIDIA AI Releases Star Elastic: One Checkpoint that Contains 30B, 23B, and 12B Reasoning Models with Zero-Shot Slicing

Learn how NVIDIA's Star Elastic technology packs multiple AI models into one file, making AI more efficient and accessible. This new method trains models of different sizes together, saving time and resources while improving performance.

May 939

research

Study shows why reasoning models often think far beyond the solution

This article explains why AI reasoning models often overthink solutions and how new research helps them know when to stop.

Feb 25110