A walkthrough of running local AI models inside the Zed editor, covering the trade-offs of local vs. cloud-hosted models (privacy, cost, availability vs. capability), how to choose a model (explaining parameters, MoE architecture, quantization), and how to configure LM Studio, Ollama, or llama.cpp as providers. Also includes practical tips for working effectively with smaller, less capable models, such as managing context windows and using subagents.
Table of contents
Why Local?How to Run Local ModelsConfiguring ZedWorking with Non-Frontier ModelsFootnotesWe're Not Building AI Features for the MoneyIntroducing Parallel Agents in ZedIntroducing Zed AISort: