← All guides

OpenCode free models

OpenCode's terminal agent can run on free models — including stealth releases like Ox Alpha. Here's where the free tiers come from, what you give up for "free," and how to get the same free-model flexibility with a real interface.

TL;DR: OpenCode is a terminal coding agent that can point at whatever model endpoint you configure, so "free models" means plugging in providers that offer a free tier — OpenRouter free models, limited-time stealth models (like Ox Alpha), or a local model you run yourself. Free tiers come with rate limits and usually data logging, so keep secrets out. If you want the same bring-any-free-model freedom but with a GUI and local-first privacy, a free open-source agent like Wuwei does that.

What "free models in OpenCode" actually means

OpenCode doesn't ship its own model — it's an agent that talks to whatever model endpoint you point it at. So running it "for free" is really about which provider you plug in. There are three common free routes, each with a different trade-off.

Ways to get free models

RouteHow it's freeThe trade-off
OpenRouter free tierSome models are offered free (rate-limited)Limits + prompts may be logged
Stealth models (e.g. Ox Alpha)Free for a limited time during testingTemporary; retains prompts
Local model (Ollama / vLLM)Runs on your own hardwareNeeds the hardware; slower on big models

The Ox Alpha example

The reason "opencode free models" spiked recently is Ox Alpha — a strong, anonymous 1M-context coding model that landed on OpenCode free for a limited time. It's a great way to see what free-tier agent coding feels like, but it's a stealth model that logs prompts, so treat it as a sandbox, not a vault.

Free almost always means "you're the training data." Free-tier and stealth models commonly log your prompts. Don't paste credentials, private keys, or proprietary code into a free hosted model. If privacy matters, run a local model or keep sensitive work on a paid, no-log provider.

Free models, but with a GUI and local privacy

If you like OpenCode's "bring any model" idea but want a graphical interface — and want your code to stay on your machine — that's exactly what a local, open-source GUI agent gives you. Wuwei is free and open source (MIT): you plug in free models (OpenRouter free tiers, stealth models, or your own local model) from a settings list, get a real interface instead of a TUI, and your code never leaves your computer.

What you get

Any free model, one list

Add OpenRouter free tiers, stealth models, or a local model and switch in a click.

Local-first

Your code stays on your machine; only the model call goes out, to the provider you choose.

Free and open source

MIT-licensed, no subscription, your keys stay yours.

FAQ

Does OpenCode have free models?

OpenCode itself is an agent, not a model — it runs on whatever endpoint you configure. You get "free models" by plugging in a provider with a free tier (like OpenRouter free models), a limited-time stealth model (like Ox Alpha), or a local model you run yourself.

Is Ox Alpha free in OpenCode?

Ox Alpha was offered free for a limited time in OpenCode. Being a free stealth model, it can log your prompts, so avoid sending anything sensitive.

How do I use free models without giving up privacy?

Run a local model (via Ollama or vLLM), or use a local-first agent so your code stays on your machine. Wuwei is a free, open-source agent that lets you plug in free or local models while keeping your code local.

Is there a free-model coding agent with a GUI?

Yes. Wuwei is a free, open-source coding agent with a graphical interface that lets you bring any model — including free tiers and local models — from a settings list.

Want OpenCode's bring-any-model flexibility, free, but with a real interface and your code kept local? Wuwei is a free, open-source AI coding agent: plug in free models, stealth models, or a local model from a list and switch in a click.

Windows · macOS · Linux — free, no login