Viability of local models for coding
Birgitta Böckeler, a colleague of Martin Fowler, spent time testing local LLMs for programming tasks. In a memo, she outlines the factors that affect how viable local models are for coding work. The article, published on Martin Fowler's blog, summarizes her findings on the practicality of running models locally versus relying on cloud-based services. The factors considered likely include model size, hardware requirements, latency, and task suitability, though specific details are not provided in the excerpt. The memo serves as a guide for developers evaluating local AI tools for their workflows.
// why it matters
Helps developers decide when local models are practical for coding tasks.