Evaluating Open-Source Local LLMs: A Comparative Analysis with Proprietary Giants for Developer Applications

This session will compare open-source local large language models (LLMs) with large proprietary models, examining their performance, flexibility, and applicability. It will explore the impact of model size on performance, the possibility of fine-tuning models for specific projects or coding languages, and the benefits of using local open-source models. The session will provide insights into selecting and deploying LLMs, with a focus on data privacy, cost-effectiveness, and customization.

Michael LundsveenØstfold University College

talkDetail.whenAndWhere

Thursday, October 10, 11:50-12:40

Room 7

talks.description

This session will evaluate open-source local large language models (LLMs) and compare their performance, flexibility, and applicability to large proprietary foundational models. We will explore how model size impacts performance, the feasibility of fine-tuning models for specific projects or coding languages, and the advantages of using local open-source models over their commercial counterparts. Attendees will gain insights into the practical considerations for selecting and deploying LLMs in development environments, focusing on data privacy, cost-effectiveness, and customization.

Data Privacy

Fine-tuning Models

Performance Comparison

Open-source LLMs

talks.speakers

Michael Lundsveen

Østfold University College

Norway

Michael is a senior engineer, researcher, and lecturer on generative AI at Østfold University College. Teaching a wide range of topics like Generative AI, Coding, Autonomous robotics, and digital fabrication. In recent years, he has focused mostly on GenAI, looking at tools, base models, and the technology's effects on society, and is investigating how technologies like large language models can be utilized in different sectors and fields.

comments.title

comments.speakerNotEnabledComments