Player not loading? Watch on YouTube
This tutorial covers installing Ollama and OpenCode, then connecting the terminal coding assistant to Ollama. The speaker recommends Ollama's own download process after reporting problems with Homebrew. The walkthrough discusses Terminal on Mac and PowerShell on Windows, and keeps the Ollama desktop app running while using the coding tools.
The OpenCode setup uses Ollama's launch integration. The speaker explains where to find installed models in the selector, how to download another model through Ollama, and how to choose a cloud model instead. Viewers who want to run models locally also get the speaker's hardware guidance: roughly 30 to 35 GB of available RAM for the 27 billion parameter model called Qwen 3.8 in the transcript. These are the speaker's recommendations, rather than measured requirements. The speaker also favors MLX models for speed on Mac but reports failures when reading images. A short DeepSeek v4 flash exchange checks that OpenCode responds.
The final section introduces ChatOSS as an open source desktop alternative to Codex that uses Ollama models. It covers opening a local project and selecting a model, followed by a Kanban workflow that the speaker says can process tasks sequentially. Orchestrate mode assigns work to subagents. The walkthrough also describes web search in the chat app and a model browser with specifications and benchmark scores.
