Model Providers and Local Models

Bring Your Own Key

OpenCode works with most major providers. Add a key for Anthropic, OpenAI or Google, and the agent uses that provider's models for its work.

Using Local Models

OpenCode supports OpenAI-compatible local servers. Tools like Ollama and LM Studio serve local models on your machine, and OpenCode can point at them instead of the cloud.

ollama serve   # start Ollama on localhost:11434

Configure the provider in the OpenCode config file, then select your local model in the session model picker.

Mixing Providers

You can use a strong cloud model for reasoning and a fast local model for simple edits, switching per session.

Key Points

• Connect any provider key to get started.
• Local models run entirely on your machine.
• Ollama and LM Studio expose OpenAI-compatible servers.
• Mix cloud and local models per session.