Gemini native format

Call , , and model listing with Google Gemini native paths and request bodies.

POST /v1beta/models/{model}:{action}

The Gemini native format keeps Google Gemini API paths and request bodies. It fits integrations that already use a Gemini SDK, a contents / parts structure, or safety settings.

The default example model is .

Available models

Model	Description
	Latest Gemini 3.6 Flash
	Gemini 3.1 Pro preview
	Gemini 3.1 Flash Lite preview
	Gemini 3 Flash preview
	Gemini 2.5 Pro
	Gemini 2.5 Flash
	Gemini 2.5 Flash Lite

Paths

Method	Path	Description
GET		List Gemini models
POST		Non-streaming content generation
POST		Streaming content generation

Replace with a concrete model ID (see Available models above), and replace with or .

The debugger on this page defaults to non-streaming .

Headers

Name	Type	Required	Description
	string	Yes	Request authentication. Use a Bearer Token, for example: . You may instead use or the query parameter
	string	No	Google API Key style auth, for example: ; mutually exclusive with
	string	Yes	Request content type; must be

The query parameter can also authenticate, for example: /v1beta/models/gemini-3.6-flash:generateContent?key=YOUR_API_KEY.

Request body

Name	Type	Required	Description
	array	Yes	Conversation contents. In multi-turn chats, alternates between and
	object	No	System-level instruction (system prompt); both and are accepted
	object	No	Generation config controlling model output behavior
	array	No	Safety settings
	array	No	Gemini tool declarations
	object	No	Tool-calling config
	string	No	Cached Content ID for reusing a context cache
	object	No	Custom key-value labels for request tracing and billing attribution

contents item

Name	Type	Required	Description
	string	No	Content role: or
	array	Yes	Content parts. Supports , ( + Base64 ), ( + ), , , and more

parts.videoMetadata

Video input metadata. Use only with video or in the same to control the time range and frame sampling rate the model reads. When there are multiple videos, each can set its own metadata.

Name	Type	Required	Description
	string	No	Video start offset as a Duration string, e.g. ,
	string	No	Video end offset, e.g.
	number	No	Frame sampling rate; default ; valid range < fps ≤

generationConfig

Name	Type	Required	Description
	number	No	Output randomness, range –, default
	number	No	Nucleus sampling probability cutoff
	integer	No	Sample only from the K highest-probability tokens
	integer	No	Number of candidate results to return; default
	integer	No	Maximum output token count
	array	No	Stop generation when any of these strings is hit; up to 5
	string	No	Output format: (default) / /
	object	No	Structured output schema (OpenAPI 3.0 subset); use with
	object	No	Standard JSON Schema constraints; mutually exclusive with
	array	No	Output modalities, e.g. ; image generation can use
	integer	No	Fixed random seed for more reproducible results
	number	No	Reduce repeated topics and encourage new content
	number	No	Reduce repeated words or sentences
	boolean	No	Whether to return token probability information
	integer	No	Number of token probabilities to return, range –
	boolean	No	Whether to enable enhanced answers for elections, government, and similar civic topics
	object	No	Speech output config (TTS); may include , , etc.
	boolean	No	Whether to return audio timestamps; for audio understanding only
	object	No	Thinking model reasoning config
	string	No	Media resolution: / /
	object	No	Image generation config; common fields include (e.g. , ) and ( / / )

thinkingConfig

Name	Type	Required	Description
	boolean	No	Whether to return a thinking summary in the response
	integer	No	Thinking token budget (Gemini 2.5 series); dynamic thinking, off
	string	No	Thinking intensity: / (Gemini 3 series; mutually exclusive with )

safetySettings item

Name	Type	Required	Description
	string	Yes	Harm category; see Common safety settings below
	string	Yes	Blocking threshold; see Common safety settings below
	string	No	Evaluation method: (by severity) or (by probability)

tools

Supported tool types:

Type	Description
	Function Calling declarations
	Google Search capability
	Code execution capability
	URL content parsing capability
	Retrieval-augmented generation (RAG); deprecated, older models only
	Google retrieval-augmented search; for new models prefer

toolConfig

Commonly used to configure :

Value	Description
	Model decides whether to call tools
	Force a tool call
	Disallow tool calls

Optional : restrict which functions may be called.

Request example

Text generation

curl -X POST https://octopusx.ai/v1beta/models/gemini-3.6-flash:generateContent \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [
      {
        "role": "user",
        "parts": [
          { "text": "Introduce the Gemini native API in three sentences." }
        ]
      }
    ],
    "generationConfig": {
      "temperature": 0.7,
      "maxOutputTokens": 300
    }
  }'

You can also use :

curl -X POST https://octopusx.ai/v1beta/models/gemini-3.6-flash:generateContent \
  -H "x-goog-api-key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [
      {
        "role": "user",
        "parts": [
          { "text": "Introduce the Gemini native API in three sentences." }
        ]
      }
    ]
  }'

Multimodal input

curl -X POST https://octopusx.ai/v1beta/models/gemini-3.6-flash:generateContent \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [
      {
        "role": "user",
        "parts": [
          {
            "inlineData": {
              "mimeType": "image/png",
              "data": "<BASE64_IMAGE>"
            }
          },
          {
            "text": "Describe the main content of this image."
          }
        ]
      }
    ]
  }'

Video input (videoMetadata)

curl -X POST https://octopusx.ai/v1beta/models/gemini-3.6-flash:generateContent \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [
      {
        "role": "user",
        "parts": [
          {
            "fileData": {
              "mimeType": "video/mp4",
              "fileUri": "https://example.com/demo.mp4"
            },
            "videoMetadata": {
              "startOffset": "3s",
              "endOffset": "10s",
              "fps": 1.0
            }
          },
          {
            "text": "Summarize the key actions in this clip."
          }
        ]
      }
    ]
  }'

Streaming generation

curl -X POST https://octopusx.ai/v1beta/models/gemini-3.6-flash:streamGenerateContent \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [
      {
        "role": "user",
        "parts": [
          { "text": "Introduce the Gemini native API in three sentences." }
        ]
      }
    ]
  }'

Response example

200 - Non-streaming

{
  "candidates": [
    {
      "content": {
        "role": "model",
        "parts": [
          {
            "text": "The Gemini native API expresses input with contents and parts. It supports text, images, files, and function calling. Through the unified gateway you can keep using the same API Key and billing system."
          }
        ]
      },
      "finishReason": "STOP",
      "safetyRatings": [
        {
          "category": "HARM_CATEGORY_HARASSMENT",
          "probability": "NEGLIGIBLE"
        }
      ]
    }
  ],
  "usageMetadata": {
    "promptTokenCount": 18,
    "candidatesTokenCount": 58,
    "totalTokenCount": 76
  }
}

Response fields

Name	Type	Description
	array	Candidate results
	object	Includes (usually ) and
	string	Finish reason, e.g.
	array	Safety ratings
	object	Token usage; includes , ,

Common safety settings

category

Value	Description
	Harassment
	Hate speech
	Sexually explicit content
	Dangerous content

threshold

Value	Description
	Do not block
	Block high risk only
	Block medium risk and above
	Block low risk and above

Related endpoints

• Text APIs overview
• Error codes