Skip to content
Console
Menu

Queues

Workflows

Getting Started

Authentication

KV Store

List AI models

Lists the admitted catalog; filter by `brand`, `modality`, or `model`.

Lists the admitted catalog; filter by brand, modality, or model.

  • Path GET https://api.sylphx.ai/v1/models — the AI gateway at api.sylphx.ai, not api.sylphx.com
  • Contract address /v1/ai_models
  • Scope ai:read
  • Effect read — nothing is written.
  • Collection ai_models
  • Query page_size, page_token, filter, order_by
  • Pagination the answer carries ai_models and, when there is more, next_page_token
  • Publishable key callable

#Request

FieldTypeWhat it is
page_sizeint32At most this many; default 50, clamped to 1000.
page_tokenstringnext_page_token of the previous page.
filterstringAIP-160 filter over labels and filterable fields.
order_bystringAIP-132 ordering over filterable fields.

#Response

FieldTypeWhat it is
ai_modelsAiModel[]The page.
next_page_tokenstringThe token of the next page; empty on the last page.

#AiModel

FieldTypeWhat it is
namestringai_models/{ai_model}.
uidstringaim_<cell><ulid>; never reused. Output only.
metaResourceMetaResource metadata.
modelstringThe canonical model id, brand/model, for example anthropic/claude-opus-5. Output only.
brandstringThe brand, for example anthropic. Output only.
aliasesstring[]Bare spellings that resolve to this model when unique, for example claude-opus-5. Output only.
modalityModelModalityWhat the model produces. Output only. One of chat, embedding, image, speech, transcription, video, decision.
input_modalitiesInputModality[]Accepted input modalities, for example text and image. Output only. One of text, image, audio, file, video.
api_surfacesApiSurface[]Inference encodings the model answers on. Output only. One of responses, messages, chat_completions, embeddings, images, speech, transcription.
context_window_tokensint64Context window in tokens. Output only.
max_output_tokensint64Most output tokens per response. Output only.
release_timetimestampThe upstream release date. Output only.
previewboolWhether the model is a preview. Output only.
pricesModelPrice[]Sell prices per meter and class; the only prices on an entry. Output only.
price_identitystringThe price identity usage events carry, for example spv3-.... Output only.
data_policyDataPolicyHow request and response data are retained and used. Output only.
cachingModelCachingHow long a prompt-cache prefix stays warm on this model, and the longest retention a request may ask for; unset when no serving lane declares it. Output only.

#ResourceMeta

FieldTypeWhat it is
generationint64Increases by one on every change to spec. Output only.
etagstringStrong ETag (AIP-154): changes on any change to spec, status, or metadata. Send it back as If-Match or etag to make Update and Delete conditional; a mismatch fails with ABORTED / 409 ETAG_MISMATCH. Output only.
create_timetimestampWhen the Resource was created. Output only.
update_timetimestampWhen the Resource last changed. Output only.
delete_timetimestampSet while the Resource is being deleted. Output only.
labelsmap<string, string>Caller-writable, indexed labels (AIP-122 label rules).
annotationsmap<string, string>Caller-writable, unindexed annotations.
display_namestringCaller-writable human-readable name.
creatorstringThe principal that created the Resource. Output only.

#ModelPrice

FieldTypeWhat it is
meterstringThe meter, for example ai.tokens. Output only.
price_classstringThe meter dimension class, for example input, cached_input, output, reasoning, cache_write (writing a prompt-cache prefix at the default retention), or cache_write_1h (writing it at the extended retention). Output only.
unitstringThe unit the price is per, for example 1M tokens or image. Output only.
currency_codestringISO 4217 currency. Output only.
unit_pricestringThe price as a decimal string, for example 1.2; never a float. Output only.

#DataPolicy

FieldTypeWhat it is
retains_promptsboolWhether Sylphx or the upstream retains prompts beyond the request. Output only.
trains_on_promptsboolWhether the upstream may train on prompts. Output only.
response_retentiondurationStored responses expire after this long (7 days). Output only.

#ModelCaching

FieldTypeWhat it is
ttldurationHow long a cached prefix stays warm after its last use with the default retention, for example 300 seconds. Output only.
extended_ttldurationThe longest retention a request may ask for with prompt_cache_retention: "24h", for example one hour on a lane whose longest native breakpoint is one hour; zero when no longer retention exists. A prefix written at this retention bills at the cache_write_1h price. Output only.

#Errors

Every error arrives in the body Errors describes.

#Examples

curl "https://api.sylphx.com/v1/ai_models" \
  -H "Authorization: Bearer $SYLPHX_API_KEY"