OpenCode Defaults to CPU Instead of GPU

[van-thieu]

Description

For local models through Ollama, OpenCode uses CPU processing instead of GPU on Linux (Ubuntu). Running Ollama standalone, it uses GPU, is there a way to force GPU usage when using command opencode serve?

Plugins

No response

OpenCode version

No response

Steps to reproduce

1. opencode serve
2. enter prompt
3. select local model through provider Ollama
4. press enter to initiate processing

OpenCode then defaults to CPU processing instead of GPU.

Screenshot and/or share link

No response

Operating System

Ubuntu 26.04

Terminal

Terminal

[github-actions]

This issue might be a duplicate of an existing issue. Please check:

• [BUG]: Ollama local models not using GPU, causing extremely slow responses #18183: Same symptom — Ollama local models not using GPU through OpenCode, causing slow responses (reported on Windows; your report is on Linux/Ubuntu)

That issue was closed. If it was not resolved for Linux, please comment there or reference it with additional context about the Ubuntu-specific case.