Model Parameters
Open the model settings panel from a chat to configure request parameters exposed by the selected cloud provider.
| Setting | Effect |
|---|---|
| Temperature | Controls sampling variability when supported. |
| Top P | Limits sampling to a cumulative probability range when supported. |
| Maximum output tokens | Caps generated output; the provider may enforce a lower limit. |
| Context limit | Informational/provider-derived budget used to validate messages and documents. |
| Image input | Enable only for a provider/model that officially supports images. |
| Native file input | Enable only for a provider/model and file type that officially supports files. |
| Tool calling | Required for MCP tools; support varies by model and API mode. |
Biyan sends only parameters accepted by the configured provider. Local-engine settings such as GPU layers, llama.cpp arguments, MLX options, and embedding mode no longer exist.
An incorrect capability flag may cause the provider to reject a request. Verify model IDs and capability support in the provider's current documentation.