Search Models Server Tool - Phaseo Docs

Add phaseo:search_models so the model can pick or recommend a model using Phaseo’s live catalogue: current modalities, providers, supported parameters, and prices.

{
  "model": "openai/gpt-6-astra",
  "messages": [
    { "role": "user", "content": "Which models can transcribe audio and return JSON?" }
  ],
  "tools": [{ "type": "phaseo:search_models", "parameters": { "max_results": 5 } }],
  "max_tool_calls": 10
}

Parameters

Parameter	Type	Default	Description
`max_results`	integer	`5`	Models returned per search, up to `20`.

What the model sends

Any of these filters. Only active models are searched, and a model must match every filter sent.

Argument	Matches
`query`	Text in the model ID, name, or description
`series`	Text in the model ID, name, or description
`input_modalities`	Models that accept all of these inputs
`output_modalities`	Models that produce all of these outputs
`provider`	A provider ID or name
`required_params`	Models that support all of these request parameters

What the tool returns

{
  "models": [
    {
      "id": "openai/gpt-4o-mini-transcribe",
      "name": "...",
      "description": "...",
      "input_modalities": ["audio"],
      "output_modalities": ["text"],
      "providers": ["openai"],
      "supported_params": ["..."],
      "pricing": {}
    }
  ],
  "total_results": 4,
  "showing": 4
}

Usage and pricing

Each search adds 1 to usage.server_tool_use.search_models_requests and can be billed with server_tool_search_models_requests.

Last modified on November 5, 2026