Published: 23 August 2024
Summary
Open-source GenAI models for coding are emerging as alternatives to proprietary models to democratize the use of AI in software engineering. This research helps software engineering leaders investigate the viability of models such as Code Llama, Codestral, CodeQwen on various code-related tasks.
Included in Full Research
Overview
Key Findings
High licensing costs and inaccessibility of most proprietary models have led organizations to explore and experiment with open-source generative AI (GenAI) code models.
Open-source GenAI models for coding may not always match the performance of proprietary solutions and often require significant infrastructure investment and extensive customization to meet specific organizational needs. This negatively impacts the overall developer experience with slower operational performance times, reduced efficiency and longer deployment times.
Open-source code models offer benefits such as community support, greater access to skilled developers, enhanced accessibility and cost savings. Yet, they also present risks similar to those of enterprise open-source
Clients can log in to view the entire
document.