Use DeepSeek, Kimi, GLM and 200+ models in OpenCode with Infravelle
Infravelle is a registered provider in OpenCode. One environment variable, 15 models out of the box, 200+ with a config line.
OpenCode is a terminal-based coding agent. Infravelle is in OpenCode's provider registry, so you get access to 15 models with one environment variable. No base URL overrides, no provider config blocks. See all available models on the models page.
Set your API key in your shell profile:
export INFRAVELLE_API_KEY="ak_live_..."
Then pass model IDs with the --model flag:
opencode run "Summarize this repo" --model infravelle/deepseek-v4-flash
opencode run "Review my code" --model infravelle/gemini-2.5-pro
Set a default model
To skip --model on every run, add an opencode.json to your project root:
{
"$schema": "https://opencode.ai/config.json",
"model": "infravelle/claude-sonnet-4-6"
}
A global default goes at ~/.config/opencode/opencode.json. Project-root config takes precedence.
Switch models mid-session
Start an interactive session with opencode, then use /models to browse and switch. All 15 registered Infravelle models appear in the picker.
Use models beyond the registry
The 15 registered models work without configuration. Infravelle serves 200+ models total. To use any of them, add a model definition to your opencode.json. The OpenCode integration guide walks through the config block.
What Infravelle does with the request
Infravelle routes each request to the cheapest available provider for that model. Routing decisions use request metadata only (token counts, context length, capability flags like streaming and tool use). We never see, log, or store your prompts, responses, or content.