stealth/ox-alpha on 2026-08-28; the OpenCode alias x-preview-f-free may stop resolving. See what the removal means.Ox Alpha in OpenCode: Setup and Fallbacks
OpenCode's live Zen catalog exposes a free model alias named x-preview-f-free. Z.ai has now confirmed Ox Alpha is a GLM-family iteration, but OpenCode still does not publish a direct Ox Alpha label for this alias, so verify the live mapping instead of hard-coding it.
1. Install the current OpenCode client
# Official installer
curl -fsSL https://opencode.ai/install | bash
# Or npm
npm i -g opencode-aiUse the official OpenCode download page for desktop builds and platform-specific instructions.
2. Connect and verify the alias
- Start OpenCode in your project.
- Run
/connectand choose the provider you intend to test. - Run
/modelsand search forx-preview-f-free. - Open the live OpenCode model list if the entry is missing or renamed.
The alias mapping is still not an exact model ID
Z.ai has confirmed the GLM-family origin, and Command Code exposes stealth/ox-alpha, while OpenCode exposes x-preview-f-free. The exact GLM-5.3-Flash checkpoint mapping is still not published by OpenCode, so verify the live provider and model label before relying on it.
3. Test with a fallback
Free previews can be capacity-limited. For a meaningful comparison, keep the same task ready with a named Z.ai model such as glm-5.3-flash or a model from the catalog. Log the client, provider, alias, latency, retries, and tool-call errors so an interruption is not mistaken for a model-quality result.
If the model stops mid-task, capture the error and retry once with a smaller continuation. If the behavior repeats, switch to the fallback rather than repeatedly resubmitting sensitive context.