Chat Completions

Generate an answer from a conversation, stream output, or request tool calls.

POST /v1/chat/completions

Request

Send JSON with Content-Type: application/json and Authorization: Bearer YOUR_BULKTOKEN_API_KEY. A minimal request contains model and messages. The example below waits for a complete JSON response.

curl https://bulktoken.ai/v1/chat/completions \
  -H "Authorization: Bearer $BULKTOKEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
  "model": "glm-5.3",
  "messages": [
    {
      "role": "user",
      "content": "Explain how a hash table works in three sentences."
    }
  ],
  "stream": false
}'

import os
from openai import OpenAI

client = OpenAI(
    api_key=os.environ["BULKTOKEN_API_KEY"],
    base_url="https://bulktoken.ai/v1",
    timeout=120.0,
)

response = client.chat.completions.create(
    model="glm-5.3",
    messages=[{"role": "user", "content": "Explain how a hash table works in three sentences."}],
    stream=False,
)
print(response.choices[0].message.content)

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.BULKTOKEN_API_KEY,
  baseURL: "https://bulktoken.ai/v1",
  timeout: 120_000,
});

const response = await client.chat.completions.create({
  model: "glm-5.3",
  messages: [{ role: "user", content: "Explain how a hash table works in three sentences." }],
  stream: false,
});
console.log(response.choices[0].message.content);

Request parameters

Field	Required	Description
`model`	Yes	Exact BulkToken model ID, visible to the API key.
`messages`	Yes	Conversation messages in order. Include relevant prior turns and tool results.
`stream`	No	Use true for Server-Sent Events, false for a single JSON response.
`stream_options`	No	Streaming options; route support varies.
`max_tokens`	No	Optional completion-token budget. Accepted limits and reasoning-token accounting depend on the model and route. Do not assume the published maximum is the default.
`temperature`	No	Optional sampling control. Accepted range, default, and compatibility with reasoning mode depend on the model.
`top_p`	No	Optional nucleus sampling control. Support and interaction with temperature depend on the model.
`stop`	No	Optional stop sequence or sequences, where supported.
`response_format`	No	Model-dependent output formatting, such as {"type":"json_object"}. Strict json_schema support must be checked per model and route.
`tools`	No	Function definitions for tool calling. Your application validates and executes requested tools.

Start with model and messages. Optional controls are model-dependent; an accepted field name does not guarantee identical behavior across every model.

Conversation messages

Each message has a role and applicable content. Use user for input and retain previous assistant messages for a multi-turn conversation. System/developer instruction handling depends on the model.

content is a string for text chat. Supported media models may also accept content parts, but accepted image/video formats and limits depend on the route. For tool calls, an assistant message can have content: null and tool_calls. Return tool results with role: "tool", the matching tool_call_id, and string content. See Tool calling.

Response

Read the answer from choices[0].message.content. A tool-call response may have no answer text. Inspect finish_reason before treating a response as complete.

The following is an illustrative response, not a record of a billable request.

{
  "id": "chatcmpl-example",
  "object": "chat.completion",
  "created": 1790553600,
  "model": "glm-5.3",
  "choices": [{
    "index": 0,
    "message": {"role": "assistant", "content": "Hello!"},
    "finish_reason": "stop"
  }],
  "usage": {"prompt_tokens": 12, "completion_tokens": 8, "total_tokens": 20}
}

Finish reasons and streaming

Reason	Meaning
`stop`	The model finished the answer or reached a stop condition.
`length`	A token budget was reached; the output can be incomplete.
`tool_calls`	Process the requested tools and return their results.
`content_filter`	The provider filtered output.

Other model-specific values can occur. With stream: true, process SSE chunks instead of this JSON response; see Streaming.

OpenAPI definition

Download the OpenAPI 3.1 definition for the documented inference and model-discovery endpoints. The request table above uses the same definition. For failures, see Errors and retries.