curl --request POST \
--url https://direct.evolink.ai/v1/responses \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data '
{
"model": "kimi-k3.1-pro",
"input": "Reply with OK.",
"reasoning": {
"effort": "low"
},
"max_output_tokens": 1024
}
'
{
"id": "resp_68f0c1c2d3e4f5a6b7c8d9e0",
"object": "response",
"created_at": 123,
"completed_at": 123,
"status": "in_progress",
"model": "<string>",
"output": [
{
"type": "reasoning",
"id": "rs_68f0c1c2d3e4f5a6b7c8d9e0",
"summary": [
{
"type": "summary_text",
"text": "<string>"
}
],
"encrypted_content": "<string>",
"status": "in_progress"
}
],
"usage": {
"input_tokens": 123,
"input_tokens_details": {
"cached_tokens": 123,
"cache_write_tokens": 123
},
"output_tokens": 123,
"output_tokens_details": {
"reasoning_tokens": 123
},
"total_tokens": 123
},
"incomplete_details": {
"reason": "max_output_tokens"
},
"error": {
"code": "<string>",
"message": "<string>"
},
"instructions": "<string>",
"reasoning": {},
"text": {},
"tools": [
{
"type": "function",
"name": "<string>",
"description": "<string>",
"parameters": {},
"strict": true
}
],
"tool_choice": "auto",
"max_output_tokens": 123,
"temperature": 123,
"top_p": 123,
"parallel_tool_calls": true,
"service_tier": "<string>",
"store": true,
"background": true,
"previous_response_id": "<string>",
"conversation": {},
"prompt_cache_options": {
"mode": "implicit",
"ttl": "5m"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}Responses API
Kimi K3.1 Pro / Highspeed - Responses API
Set reasoning effort with reasoning.effort and caching with prompt_cache_options. Search appears as web_search_call output items. For streaming, handle response.completed / response.incomplete / response.failed; hitting the output limit returns incomplete. Multi-turn requests must include the full output history, including reasoning, tool calls, and results.
POST
/
v1
/
responses
curl --request POST \
--url https://direct.evolink.ai/v1/responses \
--header 'Authorization: Bearer <token>' \
--header 'Content-Type: application/json' \
--data '
{
"model": "kimi-k3.1-pro",
"input": "Reply with OK.",
"reasoning": {
"effort": "low"
},
"max_output_tokens": 1024
}
'
{
"id": "resp_68f0c1c2d3e4f5a6b7c8d9e0",
"object": "response",
"created_at": 123,
"completed_at": 123,
"status": "in_progress",
"model": "<string>",
"output": [
{
"type": "reasoning",
"id": "rs_68f0c1c2d3e4f5a6b7c8d9e0",
"summary": [
{
"type": "summary_text",
"text": "<string>"
}
],
"encrypted_content": "<string>",
"status": "in_progress"
}
],
"usage": {
"input_tokens": 123,
"input_tokens_details": {
"cached_tokens": 123,
"cache_write_tokens": 123
},
"output_tokens": 123,
"output_tokens_details": {
"reasoning_tokens": 123
},
"total_tokens": 123
},
"incomplete_details": {
"reason": "max_output_tokens"
},
"error": {
"code": "<string>",
"message": "<string>"
},
"instructions": "<string>",
"reasoning": {},
"text": {},
"tools": [
{
"type": "function",
"name": "<string>",
"description": "<string>",
"parameters": {},
"strict": true
}
],
"tool_choice": "auto",
"max_output_tokens": 123,
"temperature": 123,
"top_p": 123,
"parallel_tool_calls": true,
"service_tier": "<string>",
"store": true,
"background": true,
"previous_response_id": "<string>",
"conversation": {},
"prompt_cache_options": {
"mode": "implicit",
"ttl": "5m"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}{
"error": {
"message": "<string>",
"type": "<string>",
"code": "<string>"
}
}BaseURL: The default BaseURL is
https://direct.evolink.ai, which has better support for text models and long-lived connections. https://api.evolink.ai is the primary endpoint for multimodal services and serves as a fallback address for text models.Authorizations
##All APIs require Bearer Token authentication##
Get API Key:
Visit the API Key Management Page to obtain your API Key
Add to request header:
Authorization: Bearer YOUR_API_KEY
Body
application/json
Use kimi-k3.1-pro or kimi-k3.1-pro-highspeed; Highspeed offers faster generation.
Available options:
kimi-k3.1-pro, kimi-k3.1-pro-highspeed input
string(Message · object | Reasoning · object | Function Call · object | Function call result · object | Custom tool call · object | Custom tool call result · object | Web search call · object | Dynamic tools · object)[]
required
Request input. A string is equivalent to one user message. An array contains typed items in order and may include conversation history, tool calls, and tool results.
Top-level system instructions, applied before all other instructions.
When true, returns the response as an SSE stream.
Maximum generated tokens, including reasoning and the final answer. Default: 131072; maximum: 1048576, also limited by the remaining context.
Required range:
1 <= x <= 1048576Reasoning configuration.
Show child attributes
Show child attributes
Output text configuration.
Show child attributes
Show child attributes
tools
(Function tool · object | Custom tool · object | Namespace tool · object | Web search tool · object)[]
Tools that the model may call.
Tool definition, distinguished by type. Supports function, custom (only apply_patch), namespace, and web_search; other tool types are unsupported.
- Function tool
- Custom tool
- Namespace tool
- Web search tool
Show child attributes
Show child attributes
Controls tool use. With auto, the model decides whether to call a tool.
Available options:
auto Additional fields to return, only for web_search. web_search_call.action.sources includes matched web sources; web_search_call.results includes image search results.
Available options:
web_search_call.results, web_search_call.action.sources Context cache-write options. If omitted, automatically writes to the 5m cache. Writes and reads are billed separately.
Show child attributes
Show child attributes
Response
Response created successfully
A model response.
Unique response identifier.
Example:
"resp_68f0c1c2d3e4f5a6b7c8d9e0"
Available options:
response Unix timestamp when the response was created.
Unix timestamp when the response ended. Present when status is completed or incomplete; null for in_progress or failed.
Response status. The initial streaming snapshot has status in_progress.
Available options:
in_progress, completed, incomplete, failed Model that generated this response.
output
(Reasoning · object | Message · object | Function Call · object | Custom tool call · object | Web search call · object)[]
Output item array, ordered as web_search_call (if any), reasoning, message, and tool calls.
An output-array element, distinguished by type.
- Reasoning
- Message
- Function Call
- Custom tool call
- Web search call
Show child attributes
Show child attributes
input_tokens includes cache reads and writes; subtract cached_tokens and cache_write_tokens to get ordinary input. output_tokens already includes reasoning.
Show child attributes
Show child attributes
Reason provided when status is incomplete.
Show child attributes
Show child attributes
Error details provided when status is failed.
Show child attributes
Show child attributes
tools
(Function tool · object | Custom tool · object | Namespace tool · object | Web search tool · object)[] | null
Tool definition, distinguished by type. Supports function, custom (only apply_patch), namespace, and web_search; other tool types are unsupported.
- Function tool
- Custom tool
- Namespace tool
- Web search tool
Show child attributes
Show child attributes
Controls tool use. With auto, the model decides whether to call a tool.
Available options:
auto Always false.
Always false.
Always null.
Always null.
Context cache-write options actually applied by the server. If prompt_cache_options was omitted, use the mode and ttl returned here.
Show child attributes
Show child attributes