Xing4.0-29B-A4B (Algorithm Code: 63650300004) — OpenAI-Compatible API Reference

API protocol: OpenAI Chat Completions
Version: Stable (max_tokens is deprecated; use max_completion_tokens)

---

1. Endpoint

Environment	URL
Public endpoint	https://api-maas.teleai.com.cn/v1/chat/completions
OpenAI-compatible base URL	https://api-maas.teleai.com.cn/v1

• Request method: POST
• Content-Type: application/json

---

2. Authentication

Use Bearer token authentication and pass your API key in the request header:

Authorization: Bearer <YOUR_API_KEY>

• For API key authentication, obtain a long-lived API key from the API Key Management page.

Header	Description
Authorization	Bearer ; required
session-id	Pass this header for multi-turn conversations. Use a different `session-id` for each session; omit it for other requests.
traceparent	W3C Trace Context header for distributed tracing, in the format {version}-{trace-id}-{parent-id}-{flags}; see the table below.

traceparent field details:

Field	Example value	Description
version	00	W3C version; default: 00
trace-id	0af7651916cd43dd8448eb211c80319c	32 hexadecimal characters; a unique ID for the entire trace
parent-id	b7ad6b7169203331	16 hexadecimal characters; the span ID for the current request
flags	01	The default 01 is sufficient

---

3. Model Capabilities and Limits

3.1 Model Capabilities

Capability	Support	Capability	Support
Input modality	Text input	Output modality	Text output
Model playground	✅ Supported	Structured Outputs	✅ Supported
Function calling	✅ Supported	Reasoning	✅ Supported
Online inference	✅ Supported	Batch inference	❌ Not supported
Explicit caching	❌ Not supported	Implicit caching	✅ Supported

3.2 Model Limits

Limit	Value	Limit	Value
Context window	256K	Maximum reasoning length	128K
Maximum input length	192K	Maximum input length (reasoning mode)	64K
Maximum output length	192K	Maximum output length (reasoning mode)	64K

---

4. Request and Response Bodies

4.1 Request Body

4.1.1 Top-Level Fields

Field	Type	Required	Description
model	string	Yes	Corresponding Model Name. Example: Xing4.0-29B-A4B
messages	array	Yes	Array of messages in conversation order; see 4.1.2
frequency_penalty	number	No	Positive values penalize new tokens based on their existing frequency in the text so far, decreasing the model's likelihood to repeat the same line verbatim; range: -2.0 to 2.0; default: 0
logprobs	boolean	No	Whether to return log probabilities for each output token; default: false
top_logprobs	integer	No	Return the top N most likely alternative tokens at each token position; range: 0 to 20; requires logprobs=true
max_tokens	integer	No	Deprecated. Use max_completion_tokens instead.
max_completion_tokens	integer	No	Maximum number of tokens generated for the completion, including reasoning tokens
n	integer	No	Number of generations to return; default: 1
presence_penalty	number	No	Positive values penalize new tokens based on whether they appear in the text so far, increasing the model’s likelihood to talk about new topics; range: -2.0 to 2.0; default: 0
repetition_penalty	number	No	Default: 1.05; range: 0 to 2. Controls repetition in generated sequences. Increasing this value can reduce repetition; 1.0 applies no penalty.
response_format	object	No	An object specifying the format that the model must output; see 4.1.3
seed	integer	No	Random seed for reproducible sampling
stop	string / array	No	Stop sequence
stream	boolean	No	Whether to stream the response; default: false
stream_options	object	No	Additional streaming options; see 4.1.4
temperature	number	No	Determines the degree of randomness in the response; range: 0 to 2; default: 1.0
top_p	number	No	An alternative to sampling with temperature, called nucleus sampling, where the model considers the results of the tokens with top_p probability mass; range: 0 to 1; default: 0.95
tools	array	No	Array of tool definitions; see 4.1.5
tool_choice	string / object	No	Tool selection behavior; see 4.1.6
parallel_tool_calls	boolean	No	Whether to allow parallel tool calls; default: true
chat_template_kwargs	object	No	Reasoning mode options. Reasoning is enabled by default; pass {"enable_thinking": true} to enable it explicitly.
reasoning_effort	string	No	Reasoning effort; this model does not support it.

Recommended sampling settings: temperature=1.0 and top_p=0.95.

4.1.2 Messages

Message object (messages[])

Field	Type	Required	Description
role	string	Yes	Role: system / user / assistant / tool
content	string / array	Yes	Message content; supports plain text or an array of content parts
name	string	No	Participant name, used to distinguish participants with the same role
tool_calls	array	No	Tool calls made by the assistant (only for assistant messages)
tool_call_id	string	No	ID of the associated tool call (required for tool messages)

Message content part (content[])

Field	Type	Required	Description
type	string	Yes	Content part type; this model supports only text
text	string	Yes	Text content

Tool call object (tool_calls[])

Field	Type	Required	Description
id	string	Yes	Tool call ID
type	string	Yes	Tool type; currently only function is supported
function	object	Yes	Tool definition

Function call object (function)

Field	Type	Required	Description
arguments	string	Yes	JSON-formatted text
name	string	Yes	Tool name

4.1.3 response_format

Structure	Description
{"type":"text"}	Default text output
{"type":"json_object"}	JSON object output (the prompt must contain the word “json”)
{"type":"json_schema","json_schema":{"name":"...","description":"...","schema":{...},"strict":true}}	Output constrained by a JSON Schema; currently unsupported

4.1.4 stream_options

Field	Type	Description
include_usage	boolean	When true, an additional chunk containing usage is returned at the end of the stream

4.1.5 Tool Definitions (tools[])

Field	Type	Required	Description
type	string	Yes	Tool type; currently function
function	object	Yes	Function definition
function.name	string	Yes	Function name
function.description	string	No	Function description
function.parameters	object	No	Parameter definition in JSON Schema format

4.1.6 tool_choice

• String: none (do not call a tool) / auto (let the model choose) / required (the model must call a tool)
• Object: {"type":"function","function":{"name":"<FUNCTION_NAME>"}} forces a specific function call

4.2 Response Body

4.2.1 Top-Level Fields

Field	Type	Description
id	string	Unique ID for this chat completion
object	string	Fixed value: chat.completion
created	integer	Creation time (Unix timestamp in seconds)
model	string	Model used for this response
choices	array	Array of completion choices
usage	object	Token usage

Completion choices (choices[])

Field	Type	Description
index	integer	Index of the choice
message	object	Message object
finish_reason	string	Reason generation stopped: stop (natural stop or stop sequence), length (limit reached), tool_calls / function_call (tool call), content_filter (content filtered), null (not yet finished), or abort (aborted).
logprobs	object	Log probabilities of output tokens (returned when logprobs is enabled)

Message object message

Field	Type	Description
role	string	Role: system / user / assistant / tool
content	string	Text content
reasoning_content/reasoning	string	Reasoning content
tool_calls[]	object	{id, type:function, function:{name, arguments}}; arguments is a JSON-formatted string

Token usage usage

Field	Type	Description
prompt_tokens	integer	Number of prompt tokens
completion_tokens	integer	Number of completion tokens
total_tokens	integer	Total number of tokens
prompt_tokens_details.cached_tokens	integer	Number of cached prompt tokens

4.2.2 Streaming (SSE) Response Fields

Field	Type	Description
id	string	Unique ID for this chat completion
object	string	Fixed value: chat.completion.chunk
created	integer	Creation time (Unix timestamp in seconds)
model	string	Model used for this response
choices	array	Array of completion choices
choices[].index	integer	Index of the choice
choices[].delta	object	Same fields as choices[].message in the response body
choices[].finish_reason	string	Reason generation stopped: stop (natural stop or stop sequence), length (limit reached), tool_calls / function_call (tool call), content_filter (content filtered), null (not yet finished), or abort (aborted).
choices[].logprobs	object	Log probabilities of output tokens (returned when logprobs is enabled)
usage	object	Token usage

5. Request Examples

5.1 curl

curl --location 'https://api-maas.teleai.com.cn/v1/chat/completions' \
--header 'Content-Type: application/json' \
--header 'Authorization: Bearer <YOUR_API_KEY>' \
--data '{
    "model": "Xing4.0-29B-A4B",
    "messages": [
        { "role": "user", "content": "Hello, introduce yourself in one sentence." }
    ],
    "stream": true
}'

5.2 Python

import requests

url = "https://api-maas.teleai.com.cn/v1/chat/completions"
api_key = "<YOUR_API_KEY>"

payload = {
    "model": "Xing4.0-29B-A4B",
    "messages": [{"role": "user", "content": "Hello, introduce yourself in one sentence."}],
    "stream": True
}

resp = requests.post(
    url,
    headers={
        "Content-Type": "application/json",
        "Authorization": f"Bearer {api_key}",
    },
    json=payload,
    stream=True,
)

# Parse the SSE stream
for line in resp.iter_lines(decode_unicode=True):
    if line and line.startswith("data:"):
        data = line[len("data:"):].strip()
        if data == "[DONE]":
            break
        print(data)

5.3 Java (with OkHttp and Jackson)

import okhttp3.*;
import com.fasterxml.jackson.databind.ObjectMapper;

public class ChatCompletionsExample {

    public static void main(String[] args) throws Exception {
        String url = "https://api-maas.teleai.com.cn/v1/chat/completions";
        String apiKey = "<YOUR_API_KEY>";

        ObjectMapper mapper = new ObjectMapper();
        String json = mapper.writeValueAsString(java.util.Map.of(
            "model", "Xing4.0-29B-A4B",
            "messages", java.util.List.of(
                java.util.Map.of("role", "user", "content", "Hello, introduce yourself in one sentence.")
            ),
            "stream", true
        ));

        OkHttpClient client = new OkHttpClient();
        Request request = new Request.Builder()
                .url(url)
                .addHeader("Content-Type", "application/json")
                .addHeader("Authorization", "Bearer " + apiKey)
                .post(RequestBody.create(json, MediaType.get("application/json")))
                .build();

        try (Response response = client.newCall(request).execute()) {
            // Non-streaming: print the response body directly
            System.out.println(response.body().string());
            // Streaming: read SSE lines from ResponseBody.source() until [DONE]
        }
    }
}