Data as of Jul 25, 2026 · Based on 2,718,867 AI responses across 9,511 prompts · See how Parse measures this
Sheared LLaMA is a family of 1.3B and 2.7B large language models produced by pruning the LLaMA2 7B model and then training on only 50B tokens. The models achieve stronger performance than existing open-source models of the same scale while using significantly less compute.
Parse Score