> ## Documentation Index
> Fetch the complete documentation index at: https://evolink.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# GPT All-Model API - Responses Reference
> - OpenAI-compatible Responses API for GPT series text models; select the specific model via `model` (see the comparison table on the `model` parameter for all allowed values)
- The whole series consists of reasoning models, with reasoning depth controlled by `reasoning.effort`; reasoning tokens are billed as output tokens
- Prompt caching applies automatically: cached input tokens are billed at the lower cached rate
- Supports both synchronous and streaming (SSE) modes
- Server-side tools: `web_search` (web search), `code_interpreter` (code execution), `file_search` (file search)
- Plain `function` tools (client-side function calls) are supported as well
- Multi-turn conversations can be chained with `previous_response_id`
- **Note** Model support varies for some parameters; see the per-parameter notes below
**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.
**Server-side tools** (`web_search`, `code_interpreter`, `file_search`, `mcp`) run on the server, so results do not need to be sent back by the client, and they are only available on this API. The Chat Completions endpoint supports regular `function` tool calling only.
**Note** This API supports synchronous and streaming modes only: the `background: true` asynchronous mode is not supported, and there are no endpoints for retrieving, cancelling or deleting a response by its ID. For long-running generations, use `stream: true` to keep the connection open.
The `image_generation` tool is not available on this model series; for image generation, use the image series model APIs.
**Multi-turn conversations**: pass the `id` returned by the previous turn as `previous_response_id` on the next turn to continue the context. Responses have a retention period; once it expires the ID is no longer valid and the request is handled as a new conversation. For scenarios with strict context-accuracy requirements, maintaining the full `input` history yourself is recommended.
## OpenAPI
````yaml en/api-manual/language-series/gpt/responses/responses-reference.json POST /v1/responses
openapi: 3.1.0
info:
title: GPT All-Model API - Responses Reference
description: >-
Full parameter reference for calling GPT series text models through the
OpenAI-compatible Responses API (including server-side tools).
license:
name: MIT
version: 1.0.0
servers:
- url: https://direct.evolink.ai
description: Production (recommended)
- url: https://api.evolink.ai
description: Alternative URL
security:
- bearerAuth: []
tags:
- name: Responses
description: OpenAI Responses API with server-side tools
paths:
/v1/responses:
post:
tags:
- Responses
summary: GPT Responses (All Models, Full Parameters)
description: >-
- OpenAI-compatible Responses API for GPT series text models; select the
specific model via `model` (see the comparison table on the `model`
parameter for all allowed values)
- The whole series consists of reasoning models, with reasoning depth
controlled by `reasoning.effort`; reasoning tokens are billed as output
tokens
- Prompt caching applies automatically: cached input tokens are billed
at the lower cached rate
- Supports both synchronous and streaming (SSE) modes
- Server-side tools: `web_search` (web search), `code_interpreter` (code
execution), `file_search` (file search)
- Plain `function` tools (client-side function calls) are supported as
well
- Multi-turn conversations can be chained with `previous_response_id`
- **Note** Model support varies for some parameters; see the
per-parameter notes below
operationId: gptResponsesReference
requestBody:
required: true
content:
application/json:
schema:
$ref: '#/components/schemas/ResponsesRequest'
responses:
'200':
description: >-
Response generated successfully (JSON object, or an SSE event stream
ending with `response.completed` when `stream=true`)
content:
application/json:
schema:
$ref: '#/components/schemas/ResponsesResponse'
'400':
description: >-
Invalid request parameters (including parameters not supported by
the model; the error message indicates the specific parameter name)
content:
application/json:
schema:
$ref: '#/components/schemas/ErrorResponse'
example:
error:
code: 400
message: >-
Invalid value: '__bogus__'. Supported values are: 'auto' and
'disabled'.
type: invalid_request_error
param: truncation
'401':
description: Unauthorized, invalid or expired token
content:
application/json:
schema:
$ref: '#/components/schemas/ErrorResponse'
example:
error:
code: 401
message: Invalid or expired token
type: authentication_error
'402':
description: Insufficient quota
content:
application/json:
schema:
$ref: '#/components/schemas/ErrorResponse'
example:
error:
code: 402
message: Insufficient quota
type: insufficient_quota_error
fallback_suggestion: https://evolink.ai/dashboard/billing
'429':
description: Rate limit exceeded
content:
application/json:
schema:
$ref: '#/components/schemas/ErrorResponse'
example:
error:
code: 429
message: Rate limit exceeded
type: rate_limit_error
fallback_suggestion: retry after 60 seconds
'500':
description: Internal server error
content:
application/json:
schema:
$ref: '#/components/schemas/ErrorResponse'
example:
error:
code: 500
message: Internal server error
type: internal_server_error
fallback_suggestion: try again later
'503':
description: Service temporarily unavailable
content:
application/json:
schema:
$ref: '#/components/schemas/ErrorResponse'
example:
error:
code: 503
message: Service temporarily unavailable
type: service_unavailable_error
fallback_suggestion: retry after 30 seconds
components:
schemas:
ResponsesRequest:
type: object
required:
- model
- input
properties:
model:
type: string
description: >-
Model to call:
| Model ID | Context window | Positioning |
|---|---|---|
| `gpt-5.6-sol` | 1,050,000 | GPT-5.6 family, frontier reasoning |
| `gpt-5.6-terra` | 1,050,000 | GPT-5.6 family, balanced production
|
| `gpt-5.6-luna` | 1,050,000 | GPT-5.6 family, high throughput and
cost control |
| `gpt-5.5` | 400,000 | General-purpose reasoning model |
| `gpt-5.4` | 128,000 | General-purpose reasoning model |
| `gpt-5.2` | 400,000 | General-purpose reasoning model |
| `gpt-5.1` | 400,000 | General-purpose reasoning model |
enum:
- gpt-5.6-sol
- gpt-5.6-terra
- gpt-5.6-luna
- gpt-5.5
- gpt-5.4
- gpt-5.2
- gpt-5.1
example: gpt-5.6-sol
input:
description: >-
Model input: a plain string, or an array of input items.
The `content` of an input item supports two block types:
`input_text` (text) and `input_image` (image):
```json
"input": [
{
"role": "user",
"content": [
{ "type": "input_text", "text": "What is in this image?" },
{
"type": "input_image",
"image_url": "https://example.com/photo.png",
"detail": "auto"
}
]
}
]
```
**Image**
- Pass the public URL of the image in `image_url`
- `image_url` must be a string; writing it as `{ "url": "..." }`
returns `400`
- `detail` is a sibling of `image_url` (not nested inside it):
`auto` (default) / `low` / `high` / `original`
- The image must be downloadable, otherwise `400` is returned
**Tool results**
- The array can also carry back tool result items from the previous
turn, such as `function_call_output`
**Note** The block types on this API differ from those on the Chat
Completions API (which uses `text` / `image_url`). They cannot be
mixed; using the wrong ones returns `400`.
oneOf:
- type: string
- type: array
items:
$ref: '#/components/schemas/InputItem'
example: >-
Search for AI news from the past week and summarize it in three
sentences.
instructions:
type: string
description: >-
System-level instructions, equivalent to inserting a system message
at the very beginning of `input`. When continuing a conversation
with `previous_response_id`, this parameter is not inherited from
the previous turn and must be passed on every turn.
example: You are a concise assistant. Answer in no more than three sentences.
stream:
type: boolean
description: >-
Whether to return a streaming response (SSE events, ending with
`response.completed`). Default `false`.
default: false
example: false
max_output_tokens:
type: integer
description: >-
Maximum number of tokens to generate (including reasoning tokens).
When the limit is reached, `status` is `incomplete`.
example: 2048
reasoning:
type: object
description: >-
Reasoning control.
**The allowed values of `effort` (reasoning depth) vary by model:**
| Model | Allowed values |
|---|---|
| `gpt-5.6-sol` / `gpt-5.6-terra` / `gpt-5.6-luna` | `none`, `low`,
`medium`, `high`, `xhigh`, `max` |
| `gpt-5.5` / `gpt-5.4` / `gpt-5.2` | `none`, `low`, `medium`,
`high`, `xhigh` |
| `gpt-5.1` | `none`, `low`, `medium`, `high` |
**`summary` (reasoning summary)**: `auto` / `concise` / `detailed`,
available across the series. Once enabled, a `reasoning` item
appears in `output`.
**`mode` (reasoning mode)**: `standard` / `pro`, supported only by
the `gpt-5.6` family.
**`context` (reasoning context scope)**: `auto` / `current_turn` /
`all_turns`, supported only by the `gpt-5.6` family.
Reasoning tokens are billed as output tokens and counted in
`usage.output_tokens_details.reasoning_tokens`.
properties:
effort:
type: string
enum:
- none
- low
- medium
- high
- xhigh
- max
example: medium
summary:
type: string
enum:
- auto
- concise
- detailed
example: auto
mode:
type: string
enum:
- standard
- pro
example: standard
context:
type: string
enum:
- auto
- current_turn
- all_turns
example: current_turn
text:
type: object
description: >-
Output text control:
- `format`: `{"type": "text"}` (default), `{"type": "json_object"}`,
or `{"type": "json_schema", "name": "...", "schema": {...},
"strict": true}` for structured output
- `verbosity`: `low` / `medium` / `high`, controls how detailed the
answer is
properties:
format:
type: object
description: Output format definition
verbosity:
type: string
enum:
- low
- medium
- high
example: medium
tools:
type: array
description: >-
Tool declarations. Server-side tools run on the server, so results
do not need to be sent back by the client:
| Tool type | Capability |
|---|---|
| `web_search` | Searches the web and browses pages (alias
`web_search_preview`) |
| `code_interpreter` | Runs code in a sandbox; requires
`"container": {"type": "auto"}` |
| `file_search` | Searches an existing vector store; requires
`vector_store_ids` |
| `mcp` | Connects to a remote MCP service; requires `server_label`
and `server_url` |
Plain `function` tools (client-side function calls) are supported as
well.
**Note** `image_generation` is not available on this model series;
use the image series model APIs instead.
items:
$ref: '#/components/schemas/Tool'
example:
- type: web_search
tool_choice:
description: >-
Controls tool selection: `"auto"` (default) / `"none"` /
`"required"`, or an object pinning a specific tool, e.g. `{"type":
"web_search"}`.
oneOf:
- type: string
enum:
- none
- auto
- required
- type: object
max_tool_calls:
type: integer
description: Maximum total number of tool calls allowed in this response.
example: 5
parallel_tool_calls:
type: boolean
description: >-
Whether the model may call multiple tools in parallel within one
turn. Defaults to `true`.
**Note** Only the `gpt-5.6` family and `gpt-5.5` support setting it
to `false`; on `gpt-5.4` / `gpt-5.2` / `gpt-5.1` this parameter has
no effect and always behaves as `true`.
default: true
example: true
previous_response_id:
type: string
description: >-
The `id` of the previous response, used to chain multi-turn
conversations without re-uploading the history.
**Note** Must be used together with `store: true` (the default).
Responses have a retention period; once it expires the ID is no
longer valid, and the request is handled as a new conversation
without inheriting context. For scenarios with strict
context-accuracy requirements, maintaining the full `input` history
yourself is recommended.
example: resp_0f5c2b2c20c39e8a006a7ef545443081979e478b10927984b5
store:
type: boolean
description: >-
Whether to store this response on the server; only stored responses
can be referenced by `previous_response_id`. Defaults to `true`.
**Note** Only the `gpt-5.6` family and `gpt-5.5` support setting it
to `false`; on `gpt-5.4` / `gpt-5.2` / `gpt-5.1` this parameter has
no effect and always behaves as `true`. If you do not want responses
stored, choose a model that supports turning it off.
default: true
example: true
include:
type: array
description: |-
Additional content to return in the response. Allowed values:
- `reasoning.encrypted_content`
- `message.output_text.logprobs`
- `web_search_call.results`
- `web_search_call.action.sources`
- `file_search_call.results`
- `code_interpreter_call.outputs`
- `message.input_image.image_url`
- `computer_call_output.output.image_url`
items:
type: string
example:
- reasoning.encrypted_content
temperature:
type: number
description: >-
Sampling temperature, ranging from 0 to 2. Lower values make the
output more deterministic.
**Note** On `gpt-5.4` / `gpt-5.2` / `gpt-5.1` the value `0` has no
effect (it is treated as unset and falls back to the default `1`);
for more deterministic output, use a value greater than 0 such as
`0.01`.
minimum: 0
maximum: 2
example: 0.7
top_p:
type: number
description: >-
Nucleus sampling parameter, ranging from 0 to 1. Adjusting it
together with `temperature` is not recommended.
minimum: 0
maximum: 1
example: 0.9
top_logprobs:
type: integer
description: >-
Number of candidate tokens returned at each position, ranging from 0
to 20; must be used together with `include:
["message.output_text.logprobs"]`.
**Note** Supported only by the `gpt-5.6` family and `gpt-5.5`; other
models do not support this parameter.
minimum: 0
maximum: 20
example: 2
frequency_penalty:
type: number
description: >-
Frequency penalty, ranging from -2 to 2, reducing the likelihood of
repeated content.
**Note** Supported only by the `gpt-5.6` family; other models do not
support this parameter.
minimum: -2
maximum: 2
example: 0.5
presence_penalty:
type: number
description: >-
Presence penalty, ranging from -2 to 2, encouraging the model to
talk about new topics.
**Note** Supported only by the `gpt-5.6` family; other models do not
support this parameter.
minimum: -2
maximum: 2
example: 0.5
truncation:
type: string
description: >-
How to handle context that exceeds the window: `disabled` (default,
returns an error) or `auto` (automatically truncates the middle of
the context).
enum:
- auto
- disabled
default: disabled
example: auto
context_management:
type: array
description: >-
Automatic compaction configuration for long conversations, for
example `[{"type": "compaction", "compact_threshold": 100000}]`: the
history is compacted automatically once the context exceeds the
threshold.
**Note** Supported only by the `gpt-5.6` family; other models do not
support this parameter.
items:
type: object
prompt_cache_key:
type: string
description: >-
Cache grouping key. Passing the same value for requests that share
the same prefix improves the prompt cache hit rate.
example: app-agent-v1
prompt_cache_retention:
type: string
description: >-
Prompt cache retention policy: `in_memory` (default) or `24h`
(extends how long the cache is retained).
enum:
- in_memory
- 24h
example: in_memory
prompt:
type: object
description: >-
References an already created prompt template, in the form `{"id":
"pmpt_xxx", "version": "1", "variables": {...}}`.
properties:
id:
type: string
version:
type: string
variables:
type: object
metadata:
type: object
description: >-
Custom key-value pairs returned as-is with the response, convenient
for tagging on the business side. Both keys and values are strings.
example:
trace_id: abc-123
safety_identifier:
type: string
description: >-
Stable identifier of the end user, used for abuse tracking.
**Note** Supported only by the `gpt-5.6` family; other models do not
support this parameter.
example: user-1024
user:
type: string
description: End-user identifier, used to distinguish the source of calls.
example: user-1024
ResponsesResponse:
type: object
properties:
id:
type: string
description: >-
Unique ID of this response, which can be used as
`previous_response_id` for the next turn
example: resp_0f5c2b2c20c39e8a006a7ef545443081979e478b10927984b5
object:
type: string
enum:
- response
description: Response type
example: response
status:
type: string
description: >-
Response status: `completed` for a normal ending, `incomplete` when
generation stopped early for reasons such as reaching
`max_output_tokens`, `failed` when generation failed
enum:
- completed
- incomplete
- failed
example: completed
model:
type: string
description: Actual model name used
example: gpt-5.6-sol
created_at:
type: integer
description: Creation timestamp
example: 1786705221
output:
type: array
description: >-
Output items in generation order: the `reasoning` item (reasoning
summary / encrypted reasoning content), tool call items (such as
`web_search_call` and `code_interpreter_call`), and finally the
`message` item containing `output_text` content.
items:
$ref: '#/components/schemas/OutputItem'
incomplete_details:
type: object
description: Explains the reason when `status` is `incomplete`
usage:
$ref: '#/components/schemas/Usage'
metadata:
type: object
description: Custom key-value pairs passed in the request, returned as-is
ErrorResponse:
type: object
properties:
error:
type: object
properties:
code:
type: integer
description: HTTP status error code
message:
type: string
description: Error description
type:
type: string
description: Error type
param:
type: string
description: Related parameter name
fallback_suggestion:
type: string
description: Suggestion when error occurs
InputItem:
type: object
description: >-
Input item: a message item (`role` + `content`), or a tool result item
from the previous turn (such as `function_call_output`)
To return a tool result:
```json
{
"type": "function_call_output",
"call_id": "call_abc123",
"output": "{\"temp_c\": 21}"
}
```
properties:
role:
type: string
description: Message role
enum:
- system
- developer
- user
- assistant
example: user
content:
description: >-
Input content: a string, or an array of content blocks (a mix of
`input_text` / `input_image`)
oneOf:
- type: string
- type: array
items:
$ref: '#/components/schemas/InputContentBlock'
type:
type: string
description: >-
Item type. Omit it for message items; use `function_call_output`
when returning a tool result.
enum:
- function_call_output
example: function_call_output
call_id:
type: string
description: >-
The `call_id` of the matching `function_call` output item from the
previous turn (not its `id`). Required only for
`function_call_output` items.
example: call_abc123
output:
type: string
description: >-
The tool execution result as a string (serialize JSON results
yourself). Required only for `function_call_output` items.
example: '{"temp_c": 21}'
Tool:
type: object
required:
- type
properties:
type:
type: string
description: Tool type
enum:
- web_search
- web_search_preview
- code_interpreter
- file_search
- mcp
- function
example: web_search
OutputItem:
type: object
properties:
id:
type: string
description: Output item identifier
type:
type: string
description: Output item type
enum:
- reasoning
- message
- web_search_call
- code_interpreter_call
- file_search_call
- mcp_call
- function_call
example: web_search_call
status:
type: string
description: Output item status
example: completed
content:
type: array
description: Message content parts (`output_text`), present on `message` items
items:
type: object
encrypted_content:
type: string
description: >-
Encrypted reasoning content, present on the `reasoning` item;
requires `include: ["reasoning.encrypted_content"]` in the request
Usage:
type: object
description: >-
Token usage statistics. Prompt caching applies automatically, and cached
input tokens are billed at the lower cached rate.
properties:
input_tokens:
type: integer
description: Number of input tokens
example: 18
output_tokens:
type: integer
description: Number of output tokens (includes reasoning tokens)
example: 42
total_tokens:
type: integer
description: Total number of tokens
example: 60
input_tokens_details:
type: object
description: Detailed input token information
properties:
cached_tokens:
type: integer
description: Number of tokens served from cache
example: 0
cache_write_tokens:
type: integer
description: Number of tokens written to cache
example: 0
output_tokens_details:
type: object
description: Detailed output token information
properties:
reasoning_tokens:
type: integer
description: Number of reasoning tokens
example: 16
InputContentBlock:
type: object
required:
- type
properties:
type:
type: string
description: |-
Content type
- `input_text`: Text block
- `input_image`: Image input
enum:
- input_text
- input_image
example: input_image
text:
type: string
description: Text content when `type=input_text`
example: What is in this image?
image_url:
type: string
description: >-
Public URL of the image (when `type=input_image`). It must be a
string; writing it as `{ "url": "..." }` returns `400`. The image
must be downloadable, otherwise `400` is returned as well.
example: https://example.com/photo.png
detail:
type: string
description: >-
Image analysis fidelity, a sibling of `image_url` (not nested inside
it)
- `low`: Low fidelity, consumes fewer tokens
- `high`: High fidelity, finer recognition
- `original`: Analyzed at the original image size
- `auto` (default): Decided automatically by the model
enum:
- auto
- low
- high
- original
default: auto
example: auto
description: >-
Multimodal content block. Declare the type via `type` and fill in only
the fields matching that type.
securitySchemes:
bearerAuth:
type: http
scheme: bearer
description: >-
##All APIs require Bearer Token authentication##
**Get API Key:**
Visit [API Key Management Page](https://evolink.ai/dashboard/keys) to
get your API Key
**Add to request header:**
```
Authorization: Bearer YOUR_API_KEY
```
````