SylphxModels
Docs menuOverview

Start

Build

Integrate

Documentation for every model we sell

One API for the whole catalog: send the official OpenAI Responses document to https://api.models.sylphx.ai/v1, authenticate with an organization key, and pay the list price published in the catalog. These pages take you from a key to a first answer, then to streaming, hosted tools, and safe retries.

First requestGet an API key
Base URLhttps://api.models.sylphx.ai/v1Every path below is relative to this.
AuthAuthorization: Bearer sk-sx-…Keys are scoped to one organization.
WireOfficial OpenAI Responses documentAlso accepted as Anthropic Messages.
Catalog27 models · 8 providersRead live from /v1/public/models.
On this pageStart here

Start hereLink to this section

Four guides cover almost every first integration. Pick the one that matches your job.

First successQuickstartMint a key, call POST /responses, and read the answer — curl, TypeScript, and Python.The wireResponses APIRequest and response shapes, streaming events, stored responses, compaction, and files.Rate sheetModels and pricingLive per-million-token prices, context windows, max output, and data posture.Model-initiatedHosted toolsLet the model search, fetch, or discover tools — and learn exactly what you observe.Failure pathsErrors and retriesOne typed envelope, idempotent creates, concrete retry rules, and rate limits.IntegratorsPlatform keysProvision machine keys for each tenant your platform hosts, then call Responses.

The whole product in 60 secondsLink to this section

Sylphx Models is a model store with one contract. Changing the model id never changes the document your client speaks.

  1. Get an API key

    Create an account on this site, then open the console and mint a key. Keys start with sk-sx-, are shown once, and are scoped to your organization. Store yours as SYLPHX_API_KEY.

  2. Send one Responses request

    Name a model from the catalog and send your input. This is the whole call — the official OpenAI Responses document, no per-model client.

    POST https://api.models.sylphx.ai/v1/responses
    curl https://api.models.sylphx.ai/v1/responses \
  -H "Authorization: Bearer $SYLPHX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"openai/gpt-5.5","input":"Hello"}'

  3. Read the answer

    The response is an official Responses object: id, output, and usage with the token counts we metered. The public model on the response is exactly the model id you sent.

    Stored by default so you can retrieve or replay it, and priced at the catalog rate for that model. See stored responses.

  4. Add what your product needs

    Turn on stream: true for SSE, list hosted tools so the model can search or fetch, and send a UUID Idempotency-Key on every retryable create.

    StreamingHosted toolsIdempotency

What you can rely onLink to this section

Four promises make the API safe to put in production. Each one links to the exact rule.

  • One terminal per stream. An SSE stream ends exactly once with a typed terminal — response.completed, response.incomplete, response.failed, or response.cancelled. EOF is not a success signal. See streaming.
  • Retries are safe with a key. Send a UUID Idempotency-Key and an identical retry replays the original response (idempotency-replayed: true) instead of running again. See retries.
  • Failures are typed. Every error is one JSON envelope with a stable code, a human summary, and a retryable flag. See the error contract.
  • Prices are the catalog. There is no second rate sheet: the catalog row for a model is what we price, and usage is metered on Responses. See models and pricing.

Endpoint mapLink to this section

Every path is relative to the base URL and authenticated with your key unless noted. The contract download lists the full machine-readable set.

MethodPathPurpose
POST/responsesCreate a response. Add stream: true for SSE.
POST/responses/compactCompact a stored response into an official compaction object.
GET/responses/{response_id}Retrieve a stored response while it is retained.
GET/responses/{response_id}/input_itemsList the input items stored for one response.
DELETE/responses/{response_id}Forget a stored response.
POST/responses/{response_id}/cancelOfficial cancel route. Responses are not background jobs.
GET/modelsThe full catalog for your key, with contract fields.
GET/models/{model}One model document for your key.
GET/public/modelsThe sell catalog without authentication.
GET/public/models/{model}One public model with its usage series.
POST/chat/completionsRetired. Returns 404 chat_completions_retired with the replacement path.

File endpoints (/files) and the second encoding (POST /messages) are described on the Responses page.

Live catalogLink to this section

Read from the public catalog when this page rendered. Prices are US dollars per million tokens; usage is what the platform metered on Responses in the last seven days.

27 models8 providersContext up to 1.05M491.9K requests · 7d
ModelInput / MOutput / MRequests · 7d
DeepSeek: DeepSeek V4.1 Flashdeepseek/deepseek-v4.1-flash$0.4725$1.89484.8K
Meta: Muse Spark 1.3 Contributormeta/muse-spark-1.3-contributor$0.126$0.2525.3K
deepseek/deepseek-flashdeepseek/deepseek-flash$0.18$0.72743
google/gemini-3.8-flashgoogle/gemini-3.8-flash$0.9$4.50643
x-ai/grok-4.6x-ai/grok-4.6$2.00$6.00357
Full rate sheetBrowse the catalog

Machine-readable contractLink to this section

The site does not re-author the wire. Import the Product Contract, generate a client, and diff it in CI.

OpenAPI/contract/ai.product.openapi.json

The download is the same schema the generated TypeScript and Rust clients build from. The contract page summarizes it in customer language first.

Contract summaryDownload OpenAPI
TipFirst success model: openai/gpt-5.5
Every sample on this site names a real catalog model. Swap the model string for any id in the catalog — the rest of your request stays identical.
Documentation · Sylphx Models