oz0 provider model
For: Operators and developers looking up a command.
Describe a model of a provider record. Aliases match an agent's needs and min_context against these, so a model the record does not describe cannot be picked for an agent that needs something. Prices are dollars per million tokens, and per second for models priced by the time they run; they win over genai-prices. --max-output is what the gateway reserves for a call to the model that sets no output limit (128000 tokens when not given).
--effort-levels and --effort-default say which reasoning effort levels it takes and its default: a job's effort is rounded to them, and a model whose levels neither its record nor the release gives is sent none. Without --shared, a shared record is copied into the tenant's own first.
Usage
oz0 provider model <provider> <model> [flags]
Flags
| Flag | Type | Default | Description |
|---|---|---|---|
--cache-read | float | dollars per million tokens read from the prompt cache | |
--cache-write | float | dollars per million tokens written to the prompt cache | |
--cache-write-long | float | dollars per million tokens written to the prompt cache with its longer lifetime | |
--capabilities | strings | what it can do: tools, vision, structured_output, thinking, pdf, audio, web_search, prompt_cache | |
--context-window | string | its context window in tokens, such as 200000, 200k or 1m | |
--effort-default | string | the effort level it runs at when a call sets none | |
--effort-levels | strings | the reasoning effort levels it takes, of low, medium, high, xhigh, max; none for a model that takes none | |
--input | float | dollars per million input tokens | |
--max-output | string | its largest output in tokens, such as 64k: what a call that sets no max_tokens is reserved for | |
--output | float | dollars per million output tokens | |
--per-second | float | dollars per second a call runs, such as a GPU's cost | |
--remove | bool | forget the model instead | |
--retires | string | the day the provider retires the model, such as 2026-12-31 (empty clears it); agents that pin it are warned before | |
--shared | bool | change the record every tenant sees (cluster admins) | |
--successor | string | the model the provider names in its place, with --retires | |
--tenant | string | default | tenant |
Inherited flags
| Flag | Type | Default | Description |
|---|---|---|---|
--context | string | $OZ0_CONTEXT or <config dir>/orchestrator-zero/context.json | CLI context file written by server init |
--log-level | string | info | log level: debug, info, warn or error |
See all oz0 commands and the global flags.