Protocols

Synthorai exposes three input protocols on the same base URL. Choose the one that matches your SDK; the gateway handles protocol translation to the upstream provider transparently.

Protocol Matrix

Endpoint	SDK	Ideal for
`POST /v1/chat/completions`	OpenAI SDK	Universal — works for every provider. e.g. `gpt-5.4-mini, qwen3.5-flash, gemini-2.5-flash, claude-sonnet-4-6`
`POST /v1/responses`	OpenAI SDK (Responses)	Tool use, reasoning, structured output. e.g. `gpt-5.4, gpt-5.4-mini, gpt-5.3-codex`
`POST /v1/messages`	Anthropic SDK	Claude-native workflows via Anthropic or Bedrock channels. e.g. `claude-sonnet-4-6, claude-opus-4-1`

Which should I use?

Already using the OpenAI SDK? Keep using it. /v1/chat/completions is the universal path and routes to any configured provider.
Already using the Anthropic SDK? Point its baseURL at the gateway and call /v1/messages. Routes to Anthropic or Bedrock-Anthropic channels.
Need tool use, reasoning, or structured output on OpenAI models? Use /v1/responses.

ℹ

All three protocols share the same API key, rate limits, billing, and quota. The only difference is request/response shape — the gateway normalizes everything internally.

Auxiliary endpoints

Beyond the three chat protocols, the gateway also proxies these OpenAI-compatible endpoints:

Parameter	Type	Description
`POST /v1/embeddings`	`embedding`	Generate vector embeddings for text.
`POST /v1/images/generations`	`image`	Generate images from a text prompt.
`POST /v1/audio/transcriptions`	`audio`	Transcribe audio to text (multipart/form-data).
`POST /v1/audio/speech`	`audio`	Convert text to speech audio.
`POST /v1/rerank`	`rerank`	Rerank documents against a query.

Streaming

Receive response tokens as they are generated using Server-Sent Events (SSE).

Enable Streaming

Set "stream": true in your request body. The response will be a stream of data: events.

curl https://synthorai.io/v1/chat/completions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-5.4-mini",
    "stream": true,
    "messages": [{"role": "user", "content": "Count to 5"}]
  }'

SSE Response Format

data: {"id":"chatcmpl-...","choices":[{"delta":{"content":"1"},"index":0}]}
data: {"id":"chatcmpl-...","choices":[{"delta":{"content":","}}]}
data: {"id":"chatcmpl-...","choices":[{"delta":{"content":"2"}}]}
data: [DONE]

ℹ

The stream ends with a data: [DONE] event. Parse each line beginning with data: and extract the choices[0].delta.content field.

All of these models — and their per-token prices — are listed on the model price comparison page.