OpenCode + Gemma 4 26B: local coding setup with Ollama

This tutorial connects OpenCode to Gemma 4 26B through Ollama on Windows. The speaker installs Ollama, sets its context value to 64K, and selects a model before installing OpenCode globally through npm. Node.js and npm must already be installed. The walkthrough then launches the coding assistant with the local model.

The speaker recommends 16 to 24GB of RAM, a four-core CPU and an Nvidia RTX GPU with at least 4GB of VRAM for partial acceleration. These are the video's recommendations, rather than verified requirements for every configuration. On an HP gaming laptop with 24GB of RAM, the speaker reports roughly 18GB of RAM and 3GB of VRAM in use, with prompt evaluation near 27 tokens per second and generation near nine tokens per second. A 4-bit model is recommended for resource usage.

The AI agent demonstration starts with requests for GPU details and hardware specifications. The speaker notes that one response reports memory incorrectly. OpenCode then runs inside the VS Code terminal to create a calculator using HTML, CSS and JavaScript. The speaker reports a build time of about 14 minutes and describes the result as functional, including dark mode, while noting minor button styling issues. The example shows a local AI coding workflow, but its performance figures and output quality reflect this particular demonstration.