Create a text completion

POST

completions

curl https://inference.tokenchange.io/v1/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $TOKENCHANGE_API_KEY" \
  -d '{
    "model": "<model-slug>",
    "stream": true,
    "max_tokens": 32,
    "prompt": "The three primary colors are"
  }'

stream = client.completions.create(
    model="<model-slug>",
    prompt="The three primary colors are",
    max_tokens=32,
    stream=True,
)

for chunk in stream:
    if chunk.choices:
        print(chunk.choices[0].text, end="")

const stream = await client.completions.create({
  model: "<model-slug>",
  prompt: "The three primary colors are",
  max_tokens: 32,
  stream: true,
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.text ?? "");
}

data: {"id":"cmpl-123","object":"text_completion","created":1767225600,"model":"<model-slug>","choices":[{"index":0,"text":" red, yellow and blue.","finish_reason":"stop"}]}

data: [DONE]

Streams a text completion for a prompt. The request and response follow the OpenAI completions format. For conversations, use chat completions.

Request

string

required

Bearer ask_..., an inference key.

string

required

A model slug from the catalog, or <model-slug>/ep/<endpoint> to pin one endpoint.

string

required

The text to complete.

boolean

required

Must be true.

integer

Maximum tokens to generate. At least 1.

number

Sampling temperature, from 0 to 2.

boolean

default:"false"

Set to true to receive token usage in the final chunk.

string

Requests with the same key are routed to the same endpoint. See Prompt caching.

object

Price options. See Price options.

Other OpenAI completion fields, such as top_p, stop and seed, are passed to the endpoint unchanged.

curl https://inference.tokenchange.io/v1/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $TOKENCHANGE_API_KEY" \
  -d '{
    "model": "<model-slug>",
    "stream": true,
    "max_tokens": 32,
    "prompt": "The three primary colors are"
  }'

stream = client.completions.create(
    model="<model-slug>",
    prompt="The three primary colors are",
    max_tokens=32,
    stream=True,
)

for chunk in stream:
    if chunk.choices:
        print(chunk.choices[0].text, end="")

const stream = await client.completions.create({
  model: "<model-slug>",
  prompt: "The three primary colors are",
  max_tokens: 32,
  stream: true,
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.text ?? "");
}

Response

A stream of server-sent events. Each event’s data is a JSON completion chunk with object set to text_completion, and the stream ends with data: [DONE]. Each choice carries the new text and, in its last chunk, a finish_reason. Usage works as in chat completions.

data: {"id":"cmpl-123","object":"text_completion","created":1767225600,"model":"<model-slug>","choices":[{"index":0,"text":" red, yellow and blue.","finish_reason":"stop"}]}

data: [DONE]

The values above are illustrative.

⌘I

curl https://inference.tokenchange.io/v1/completions \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $TOKENCHANGE_API_KEY" \
  -d '{
    "model": "<model-slug>",
    "stream": true,
    "max_tokens": 32,
    "prompt": "The three primary colors are"
  }'

stream = client.completions.create(
    model="<model-slug>",
    prompt="The three primary colors are",
    max_tokens=32,
    stream=True,
)

for chunk in stream:
    if chunk.choices:
        print(chunk.choices[0].text, end="")

const stream = await client.completions.create({
  model: "<model-slug>",
  prompt: "The three primary colors are",
  max_tokens: 32,
  stream: true,
});

for await (const chunk of stream) {
  process.stdout.write(chunk.choices[0]?.text ?? "");
}

data: {"id":"cmpl-123","object":"text_completion","created":1767225600,"model":"<model-slug>","choices":[{"index":0,"text":" red, yellow and blue.","finish_reason":"stop"}]}

data: [DONE]

​Request

​Response

Request

Response