vetta models

vetta models list is how you learn a legal --model for vetta agent create, which requires one. There is no fixed list. The catalogue is read live from the network, so a model published today is runnable today — and hundreds are catalogued, which is why the listing pages and searches.

Commands

Command	Description
`vetta models list`	List available models, optionally narrowed to one completion window.
`vetta models get <id>`	One model by id, for the id you already hold.

list

vetta models list --search glm --limit 2

{
  "data": [
    {
      "object": "model",
      "id": "zai-org/GLM-5.2-FP8",
      "context_window": 1048576,
      "max_output_tokens": 32768,
      "supported_windows": ["immediate", "priority", "loose"],
      "efforts": ["low", "medium", "high"]
    },
    {
      "object": "model",
      "id": "moonshotai/Kimi-K2.6",
      "context_window": 262144,
      "max_output_tokens": 16384,
      "supported_windows": ["immediate", "priority", "loose"],
      "efforts": ["low", "medium", "high"]
    }
  ],
  "has_more": true,
  "next_cursor": "2"
}

Flag	Description
`--window`	Narrow to models servable in one completion window: `immediate` \| `priority` \| `loose`.
`--search`	Matches a model’s id and name.
`--limit`	Page size, 1–100. Defaults to 20.
`--after`	The `next_cursor` of the previous page.

context_window is in tokens. max_output_tokens is the longest reply that model may produce, and what the pre-flight quote is bounded by. supported_windows is the authoritative list of windows that model can actually be served in — an agent create naming a window outside it is rejected. No price is on the reply. What an agent spends is bounded by its own budget, and read back with vetta agent spend. One id in the listing is not a specific model: vetta/auto picks one per request. It serves immediate only — see Vetta Auto.

get

vetta models get zai-org/GLM-5.2-FP8

{
  "object": "model",
  "id": "zai-org/GLM-5.2-FP8",
  "context_window": 1048576,
  "max_output_tokens": 32768,
  "supported_windows": ["immediate", "priority", "loose"],
  "efforts": ["low", "medium", "high"]
}

An id this deploy does not serve is a not_found error, so this is also how you check a model id before you run an agent on it.

Not every model serves every window

immediate is served by every catalogued model. priority and loose are served only by the models the completion-window pool hosts, and supported_windows on each row is the authoritative answer — it is derived from the rate card the router prices against, so a model listed for a window can never be refused for it.

vetta models list --window loose

{
  "data": [
    {
      "object": "model",
      "id": "google/gemma-4-31B-it",
      "context_window": 131072,
      "max_output_tokens": 16384,
      "supported_windows": ["immediate", "priority", "loose"],
      "efforts": []
    }
  ],
  "has_more": true,
  "next_cursor": "1"
}

Naming a window a model does not serve is refused with window_unavailable (HTTP 400) before any inference runs — never silently downgraded to a dearer window. So list first, then create.

Picking a model in a script

vetta models list --window immediate --limit 1 | jq -r '.data[0].id'

Getting started

Building & running

Organization

Identity

vetta models

Commands

list

get

Not every model serves every window

Picking a model in a script

​Commands

​list

​get

​Not every model serves every window

​Picking a model in a script

Commands

list

get

Not every model serves every window

Picking a model in a script