GitHub Copilot CLI now supports local LLM models, making GitHub authentication optional and enabling offline/air-gap usage. Users can connect to any OpenAI-compatible endpoint — including Ollama, Llama.cpp, or other local model runners — by setting environment variables for base URL, API key, and model name. A demo shows Gemma 4 running via Llama.cpp in Docker on localhost:8080 connected to Copilot CLI. The author views this as a positive move by GitHub, contrasting it with Claude Code's undocumented base URL workaround, which they consider unreliable. They still recommend OpenCode for local model use due to its community support and better tool execution compatibility.
•2m watch time
1 Impression