Models and Network & capacity
Models lists every model this workspace can call, its supply under your requirements, and a path straight into the Playground. Its filters follow your key's requirements, so what you see is what your requests would get.
Shown per pool; under Auto, each request runs in whichever pool serves it best on price and speed, under the access conditions of the pool that serves it
| State | What it means |
|---|---|
| Available | The pool can serve the model under these requirements now |
| Limited | It can, with less capacity |
| Unavailable | It cannot under these requirements. The Console never relaxes a requirement on your behalf |
| Unknown | No reading yet |
- State
- Available
- What it means
- The pool can serve the model under these requirements now
- State
- Limited
- What it means
- It can, with less capacity
- State
- Unavailable
- What it means
- It cannot under these requirements. The Console never relaxes a requirement on your behalf
- State
- Unknown
- What it means
- No reading yet
While the workspace excludes the Community pool, Auto runs in the Certified pool alone, with fallback off, and the screen says so beside the pool.
| The model page says | What it means |
|---|---|
| Validated | A real tool call passed this model's validation. Requests with tools are taken |
| Supported by its template, not tested yet | Its chat template renders tools and tool calls and the engine reads the model's tool calls through it, so requests with tools are taken; no validation has run a real tool call on it yet |
| No tool calls | Requests with tools are refused with 400 unsupported_parameter |
| Not read yet, or not reported | Nothing is claimed either way yet |
- The model page says
- Validated
- What it means
- A real tool call passed this model's validation. Requests with tools are taken
- The model page says
- Supported by its template, not tested yet
- What it means
- Its chat template renders tools and tool calls and the engine reads the model's tool calls through it, so requests with tools are taken; no validation has run a real tool call on it yet
- The model page says
- No tool calls
- What it means
- Requests with tools are refused with 400
unsupported_parameter
- The model page says
- Not read yet, or not reported
- What it means
- Nothing is claimed either way yet
For each coding agent HIPS tests (HIPS Code, OpenCode, Claude Code and Codex), a model shows whether a test of the agent passed on its current version, with the agent's version, the pool and the date; whether it meets what the agent needs and is not tested yet; or why it does not (no tool calls, or a context below the least the agent works in). A failed test names what went wrong.
The routers are listed first, as a group of their own: name one as the model and HIPS chooses the model for each request. The Router has the rest.
Each model shows its rates per pool, input and output per 1M tokens for a language model, with their history. A request pays the rate in force when it is accepted; Billing and credits has how a request is priced.
A model's page has its inputs and outputs, its context window and limits, its license, its supply and rates per pool, its tool calls and coding agents, and the code to call it, with the model id filled in and the key read from HIPS_API_KEY.
Network & capacity shows the supply as the network publishes it, per pool, with when each reading was taken, and the always-on configurations of the Certified pool: the GPU type and count, memory, the region the configuration runs in, availability and the hourly quote. A capacity snapshot is not a reservation. Always-on capacity comes from the Certified pool only.
Related sheets
- Sheet
- 11 of 19
- Detail
- F
- Updated
- 2026-11-11