Use MagmaRouter in Zed

Zed's Agent Panel supports an openai_compatible provider. Declare it in settings.json with the model list; the key is entered through the UI so it stays in the keychain.
- Base URL
- https://api.magmarouter.com/v1
- API key
- Get one, starts sk-mr-
- Example model
- glm-5.3-xploded
An openai_compatible provider
In settings.json:
JSON
{
"language_models": {
"openai_compatible": {
"magmarouter": {
"api_url": "https://api.magmarouter.com/v1",
"available_models": [
{
"name": "glm-5.3-xploded",
"display_name": "GLM 5.3 Flash Xploded (MagmaRouter)",
"max_tokens": 1048576
}
]
}
}
}
}Add the key through Agent Settings, "Add Provider" (it is stored in your OS keychain, never in the settings file). Each model's capabilities.tools defaults to true; set it to false for a non-tool model. Zed's inline edit predictions need a legacy completions endpoint MagmaRouter does not serve, so use MagmaRouter for the Agent Panel chat and edits, not for edit predictions.
Pick a model
Each page has the Zed setup with that model's id and limits filled in, and what a session costs.
GLM 5.3 Xploded in Zed1M context, $2.00 in / $5.00 out per 1M
GLM 5.3 Flash Xploded in Zed1M context, $0.30 in / $1.00 out per 1M
DeepSeek V4.1 Flash 1017 Xploded in Zed1M context, $1.25 in / $2.50 out per 1M
DeepSeek Flash Lite Xploded in Zed512K context, $0.36 in / $1.44 out per 1M
MiMo V2.6 Flash Xploded in Zed1M context, $1.00 in / $2.00 out per 1M
Qwen3.8 27B Xploded in Zed1M context, $1.50 in / $5.00 out per 1M
Qwen3.8 27B Ultrafast Xploded in Zed524K context, $1.75 in / $5.50 out per 1M