ResolvedModelCapabilities

Normalized runtime behavior that UI and API consumers can safely inspect.

familyFamily (string)

Resolved model family used for runtime behavior decisions.

Default value:

supports_thinkingSupports Thinking (boolean)

Whether the runtime expects the model to expose a reasoning or thinking mode.

Default value: false

supports_thinking_toggleSupports Thinking Toggle (boolean)

Whether thinking can be explicitly enabled or disabled for requests.

Default value: false

supports_thinking_budgetSupports Thinking Budget (boolean)

Whether the runtime expects the model to accept a thinking or reasoning budget control.

Default value: false

default_reasoning_effortDefault Reasoning Effort (string)

Reasoning effort used when thinking is enabled without an explicit effort override.

Possible values: [none, minimal, low, medium, high, xhigh]

Default value: medium

disabled_reasoning_effortDisabled Reasoning Effort (string)

Reasoning effort used when thinking is explicitly disabled.

Possible values: [none, minimal, low, medium, high, xhigh]

Default value: none

thinking_formatThinking Format (string)

Resolved reasoning marker format expected from this model family.

Default value: none

supports_image_inputSupports Image Input (boolean)

Whether the runtime should treat the model as accepting image inputs.

Default value: false

supports_audio_inputSupports Audio Input (boolean)

Whether the runtime should treat the model as accepting audio inputs.

Default value: false

supports_speech_synthesisSupports Speech Synthesis (boolean)

Whether the runtime should treat the model as a text-to-speech model.

Default value: false

supports_transcriptionSupports Transcription (boolean)

Whether the runtime should treat the model as a speech-to-text model.

Default value: false

supports_speech_translationSupports Speech Translation (boolean)

Whether the runtime should treat the model as supporting speech translation.

Default value: false

supports_audio_outputSupports Audio Output (boolean)

Whether the runtime should expect this model to produce audio output.

Default value: false

supports_realtime_audioSupports Realtime Audio (boolean)

Whether the runtime should expect this model to expose realtime audio sessions.

Default value: false

default_audio_response_format object

Default encoded audio response format for speech synthesis, when declared.

anyOf

• string
• null

string

audio_response_formatsstring[]

Encoded audio response formats the model can produce.

supports_tool_callingSupports Tool Calling (boolean)

Whether the runtime expects the model to support structured tool calling.

Default value: false

builtin_toolsstring[]

Builtin platform tool contracts that Skulk may expose to this model family.

tool_call_formatTool Call Format (string)

Resolved tool-call output format family used for parsing.

Default value: generic

prompt_rendererPrompt Renderer (string)

Resolved prompt renderer strategy used to prepare requests for this model.

Default value: tokenizer

output_parserOutput Parser (string)

Resolved output parser strategy used to interpret model responses.

Default value: generic

supports_native_multimodalSupports Native Multimodal (boolean)

Whether the runtime can use a native multimodal execution path for the model.

Default value: false

ResolvedModelCapabilities

{
  "family": "",
  "supports_thinking": false,
  "supports_thinking_toggle": false,
  "supports_thinking_budget": false,
  "default_reasoning_effort": "medium",
  "disabled_reasoning_effort": "none",
  "thinking_format": "none",
  "supports_image_input": false,
  "supports_audio_input": false,
  "supports_speech_synthesis": false,
  "supports_transcription": false,
  "supports_speech_translation": false,
  "supports_audio_output": false,
  "supports_realtime_audio": false,
  "default_audio_response_format": "string",
  "audio_response_formats": [
    "string"
  ],
  "supports_tool_calling": false,
  "builtin_tools": [
    "string"
  ],
  "tool_call_format": "generic",
  "prompt_renderer": "tokenizer",
  "output_parser": "generic",
  "supports_native_multimodal": false
}