Using TAO models

The built-in tao provider connects taoCode to Bittensor inference gateways. Connect a gateway, then choose a model from the list available to your account.

Supported gateways

Gateway	Subnet	Environment variable
Chutes	SN64	CHUTES_API_KEY
Targon	SN4	TARGON_API_KEY

See Gateways for account links and connection details.

List available TAO models

taocode models

On macOS or Linux, filter the output with taocode models | grep 'tao/'. In PowerShell, use taocode models | Select-String 'tao/'.

Browse the model catalog to compare capabilities, then use the CLI list to see the choices available through your connected gateways.

Read a model reference

tao / chutes / organization/model-id
 │      │         └─ model name or alias
 │      └─────────── gateway
 └────────────────── provider

Pass the full reference to --model. A model with a similar name on another gateway is a different endpoint and may have different capabilities or limits.

Key resolution order

For each gateway, taoCode checks stored credentials (tao-chutes or tao-targon in auth.json) before its environment variable. A stale stored key can therefore mask a replacement key exported in your shell.

Use /connect to update the stored credential. With no gateway credentials, TAO models will not be available through authenticated gateways.

Choose the next step

Browse the model catalog, learn how to select a model, or follow the request lifecycle.