Apertus LLM Family Expands with Distillation and Quantization

Researchers have developed a method to expand the Apertus LLM family by creating smaller, more efficient models through distillation and quantization. The new Apertus-v1.1 models, with up to 4B parameters, were trained on 1.7T tokens and demonstrate cost-effectiveness while maintaining strong accuracy. This approach allows LLMs to be adapted for a wider range of hardware and budget constraints, making them suitable for diverse applications. AI

IMPACT Enables LLMs to meet diverse hardware and budget constraints, broadening their applicability across various use cases.

RANK_REASON The cluster describes a research paper detailing the creation of new LLM variants through distillation and quantization. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• Apertus LLM
• Apertus-v1.1

• paper
• model release

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →