Catalog models
Discover published public catalog models and their served capabilities.
GET /v1/catalog/models
Authorization
Authorizationheader · string- Public. No API key is required; supplied credentials do not change the catalog.
Request
Request · cURL
curl "https://api.boardwalk.cloud/v1/catalog/models"Parameters
This endpoint accepts no request body or query parameters.
Response
Example response. IDs, token counts, and rates are illustrative.
200 · application/json
{
"object": "list",
"data": [
{
"id": "Qwen/Qwen3-0.6B",
"object": "model",
"created": 1789516800,
"owned_by": "boardwalk",
"display_name": "Qwen3 0.6B",
"hf_repo": "Qwen/Qwen3-0.6B",
"hf_revision": "0123456789abcdef0123456789abcdef01234567",
"config_revision": "cfg_example",
"dtype": "bfloat16",
"quantization": "none",
"max_model_len": 32768,
"supports_reasoning": true,
"engine": "vllm",
"engine_version": "example",
"price_in_micros_per_mtok": "100000",
"price_out_micros_per_mtok": "300000",
"price_cache_read_micros_per_mtok": "50000",
"price_cache_write_micros_per_mtok": "0",
"status": "active",
"input_modalities": ["text"],
"output_modalities": ["text"],
"supports_tool_calls": true,
"supports_json_mode": true,
"supports_structured_output": true,
"supports_image_input": false,
"supports_image_output": false,
"supports_streaming": true,
"context_window": 32768,
"max_output_tokens": null,
"supported_endpoints": [
"/v1/chat/completions",
"/v1/completions",
"/v1/tokenize"
],
"supported_parameters": [
"model",
"stream",
"stream_options",
"max_tokens",
"temperature",
"top_p",
"n",
"stop",
"presence_penalty",
"frequency_penalty",
"seed",
"logprobs",
"top_logprobs",
"logit_bias",
"user",
"messages",
"max_completion_tokens",
"tools",
"tool_choice",
"parallel_tool_calls",
"response_format"
],
"supported_response_formats": ["text", "json_object", "json_schema"],
"capabilities_verified_at": 1789516800
}
]
}object"list"- Object type.
dataobject[]- Published catalog models with per-token rates. Your own deployments are listed separately by GET /v1/models.
Child fields
idstring- Callable model ID. Pass this as model in an inference request.
object"model"- Object type.
createdinteger- Creation time in Unix seconds.
owned_bystring- Boardwalk for public models, or the owning organization slug.
display_namestring- Display name.
hf_repostring- Source Hugging Face repository.
hf_revisionstring- Pinned weight commit.
config_revisionstring- Serving configuration revision.
dtypestring- Serving precision.
quantizationstring- Quantization method, or none.
max_model_leninteger- Maximum combined input and output tokens.
supports_reasoningboolean- Whether the active serving configuration has verified reasoning output.
input_modalitiesstring[]- Accepted input types. Qualified vision revisions expose ["text", "image"]; inspect image_input for limits. Audio and video are unsupported.
output_modalitiesstring[]- Generated output types. Currently ["text"]; image, audio, and video generation are not supported.
supports_tool_callsboolean- Whether the served revision parses tool calls: its pinned chat template names a tool-call format.
supports_json_modeboolean- Whether JSON mode is enforced by the served configuration's engine. Check formats for the endpoint.
supports_structured_outputboolean- Whether JSON Schema output is enforced by the served configuration's engine. Check formats for the endpoint.
supports_image_inputboolean- Whether this served revision accepts JPEG/PNG images. Requires qualified image processing, token accounting and pricing.
image_inputobject | null- Image limits and processor_revision, or null for text-only models. The initial profile accepts two JPEG/PNG images, each up to 3 MiB, 4096 pixels per side and 4,194,304 pixels total. Sources are inline base64 or caller-hosted HTTPS URLs. detail may be omitted or auto; low/high are rejected. No uploaded files or file IDs.
supports_image_outputboolean- Whether this endpoint can generate images. Currently false.
supports_streamingboolean- Whether the served configuration streams, with usage and a terminal event.
context_windowinteger- Shared input/output token ceiling, also exposed as max_model_len.
supported_parameters_by_endpointobject- API path to supported top-level field names. Missing endpoints are unavailable. Field values remain subject to documented constraints.
supported_response_formats_by_endpointobject- Generation API path to output format names. Messages has no json_object mode.
reasoning_controlsobject- thinking_toggle, effort_values, and token_budget. Current eligible Qwen3 templates accept effort none; reasoning output alone does not imply a control.
tool_callingobject- Detailed Chat support: tool_choice (auto, required, named, none), parallel_calls, single_call, strict_arguments, tool_results. Anything false is refused. strict_arguments covers tool_choice auto; strict always holds with required or a named tool. Native routes have additional restrictions.
capability_sourcesobject- Feature name to its source: model_metadata (the pinned chat template) or engine (the reviewed engine contract).
max_output_tokensinteger | null- Separate model output ceiling, when established. Null is not unlimited: input and output still share context_window.
supported_endpointsstring[]- Available paths for this serving configuration. Native Messages and Responses depend on how the revision is served; read this list rather than assuming them.
supported_parametersstring[]- Top-level Chat Completions fields available for this revision. See supported_response_formats for response_format values. Reasoning output does not imply reasoning_effort controls.
supported_response_formatsstring[]- Chat response_format.type values the engine contract establishes. Use the endpoint-specific map for native APIs.
capabilities_verified_atinteger | null- When the serving revision's deploy finished its checks, in Unix seconds. Null means no deploy recorded one.
enginestring- Serving engine.
engine_versionstring- Pinned engine version.
price_in_micros_per_mtokstring- Uncached input rate. Integer micro-dollars per million tokens. Zero for your own deployments, which bill GPU time.
price_out_micros_per_mtokstring- Output rate. Integer micro-dollars per million tokens.
price_cache_read_micros_per_mtokstring- Cache-read rate. Integer micro-dollars per million tokens.
price_cache_write_micros_per_mtokstring- Cache-write rate. Integer micro-dollars per million tokens.
pricingobject- The uncached input and output rates as US dollars per million tokens: { input, output }. The micro-dollar fields above are authoritative; this restates them in the units partner catalogs read.
context_lengthinteger- Context window in tokens, under the name partner catalogs read. Equals max_model_len.
statusstring- Current model status.
Capability results
Capabilities describe this served revision through Boardwalk's API. True means we serve the feature; false means a request using it is refused before it reaches a GPU. capability_sources says where each comes from: the revision's pinned chat template, or the reviewed engine contract. They describe what our serving path applies, not how well a model uses it.
Use supported_parameters_by_endpoint and supported_response_formats_by_endpoint for the route you call. The older unqualified lists describe Chat. JSON and schema support booleans summarize all available routes, so check the endpoint maps before sending a request. Parameters are top-level names; value restrictions still apply. context_window is the shared input/output ceiling. max_output_tokens: null means no separate output ceiling is established, not unlimited generation.
JSON mode produces a JSON object. Structured output additionally enforces a JSON schema. Reasoning support means reasoning is returned separately from the answer; it does not imply support for reasoning effort or budget controls.
Errors
Errors use the shared error format. Save the x-request-id response header for receipts and support.