Gemini (Google Protocol)

1. Overview

A multimodal AI model from Google, designed to handle various data types including text, images, audio, video, and code.

API Compliance

This API is compliant with the Google Gemini interface specification and supports all official parameters.

Documentation Reference

This document lists only a subset of parameters. For a complete list, please refer to the official documentation.

Supported Google Models

gemini-3.1-pro-preview (Newest)
gemini-3.1-flash-lite-preview
gemini-3-flash-preview
gemini-3-pro-preview (Deprecated, migrated to gemini-3.1-pro-preview)
gemini-2.5-flash (Planned for deprecation in Oct 2026)
gemini-2.5-pro (Planned for deprecation in Oct 2026)
gemini-2.5-flash-lite

2. Request Instructions

Request Method: POST
Request URL: https://api.exchangetoken.ai/v1beta/models/{model}:generateContent
Request URL (Streaming): https://api.exchangetoken.ai/v1beta/models/{model}:streamGenerateContent?alt=sse

3. Request Parameters

3.1 Header Parameters

Parameter	Type	Required	Description	Example
Content-Type	string	Yes	The request content type. Must be `application/json`.	`application/json`
Accept	string	Yes	The expected response content type. It's recommended to use `application/json`.	`application/json`
Authorization	string	Yes	The API Key for authentication. Format: `Bearer $YOUR_API_KEY`.	`Bearer $YOUR_API_KEY`

3.2 Body Parameters (application/json)

Parameter	Type	Required	Description	Example
contents	array	Yes	The content of the conversation with the model. For a single-turn query, this is a single instance. For multi-turn queries (like chat), this is a repeated field containing the conversation history.	`[{"role":"user", "parts":[{"text":"A cute baby sea otter"}]}]`
content.role	string	Yes	The role of the message author. Must be `user` or `model`.	`user`
content.parts	array	No	An ordered list of parts that make up a single message. Parts can have different MIME types.	`[{"text":"A cute baby sea otter"}]`
content.parts.text	string	No	Inline text content.	`A cute baby sea otter`
content.parts.inlineData	struct	No	Inline media bytes.	-
content.parts.inlineData.mimeType	string	Yes	The IANA standard MIME type of the source data.	`image/png`
content.parts.inlineData.data	string	Yes	The raw bytes of the media format. A base64-encoded string.	-
generationConfig	struct	No	Configuration options for model generation and output.	-

Complete Code Examples

1. Text Generation

cURLPython

bash

curl "https://api.exchangetoken.ai/v1beta/models/gemini-3-flash-preview:generateContent" \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -X POST \
  -d '{
    "contents": [
      {
        "role": "user",
        "parts": [
          {
            "text": "How does AI work?"
          }
        ]
      }
    ]
  }'

python

import requests
import json

url = "https://api.exchangetoken.ai/v1beta/models/gemini-3-flash-preview:generateContent"
headers = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json"
}

data = {
    "contents": [
        {
            "role": "user",
            "parts": [
                {"text": "How does AI work?"}
            ]
        }
    ]
}

response = requests.post(url, headers=headers, json=data)
print(response.json())

2. Thinking Model

cURLPython

bash

curl "https://api.exchangetoken.ai/v1beta/models/gemini-3-flash-preview:generateContent" \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -X POST \
  -d '{
    "contents": [
      {
        "role": "user",
        "parts": [
          {
            "text": "How does AI work?"
          }
        ]
      }
    ],
    "generationConfig": {
      "thinkingConfig": {
        "thinkingLevel": "low"
      }
    }
  }'

python

import requests
import json

url = "https://api.exchangetoken.ai/v1beta/models/gemini-3-flash-preview:generateContent"
headers = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json"
}

data = {
    "contents": [
        {
            "role": "user",
            "parts": [
                {"text": "How does AI work?"}
            ]
        }
    ],
    "generationConfig": {
        "thinkingConfig": {
            "thinkingLevel": "low"
        }
    }
}

response = requests.post(url, headers=headers, json=data)
print(response.json())

Controlling Thinking

The Gemini model uses dynamic thinking by default, automatically adjusting its reasoning effort based on the complexity of the user's request. However, if you have specific latency constraints or need the model to engage in deeper reasoning, you can use parameters to control its thinking behavior.

Thinking Level (Gemini 3)

The thinkingLevel parameter (recommended for Gemini 3 and later models) can be used to control reasoning behavior.

The following table details the thinkingLevel settings for each model type:

Thinking Level	Gemini 3 Pro	Gemini 3 Flash	Description
minimal	Not Supported	Supported	Matches the "no thinking" setting for most queries. For complex coding tasks, the model may perform very simple thinking. Minimizes latency for chat or high-throughput applications. Note that `minimal` does not guarantee that thinking is turned off.
low	Supported	Supported	Minimizes latency and cost. Best for simple instruction following, chat, or high-throughput applications.
medium	Not Supported	Supported	Balanced thinking, suitable for most tasks.
high	Supported (Default, Dynamic)	Supported (Default, Dynamic)	Maximizes reasoning depth. The model may take longer to generate the first (non-thinking) output token, but the result will be more carefully reasoned.

For Gemini 3 Pro, you cannot disable thinking. Gemini 3 Flash also doesn't support completely turning off thinking, but the minimal setting implies the model might not think (though it still could). If you don't specify a thinking level, Gemini will use the default dynamic thinking level of "high" for Gemini 3 models.

Gemini 2.5 series models do not support thinkingLevel; use thinkingBudget instead.

Thinking Budget

The thinkingBudget parameter, introduced with the Gemini 2.5 series, provides guidance to the model on the specific number of thinking tokens to use for reasoning.

Important Note

Use the thinkingLevel parameter with Gemini 3 models. While thinkingBudget can be used for backward compatibility, using it with Gemini 3 Pro may result in suboptimal performance.

Below are the details for the thinkingBudget configuration for each model type. You can disable thinking by setting thinkingBudget to 0. Setting it to -1 enables dynamic thinking, meaning the model adjusts the budget based on the request's complexity.

Model	Default Setting (no budget set)	Range	Disable Thinking	Enable Dynamic Thinking
2.5 Pro	Dynamic Thinking	128 to 32768	Not applicable: Thinking cannot be disabled	`thinkingBudget = -1` (Default)
2.5 Flash	Dynamic Thinking	0 to 24576	`thinkingBudget = 0`	`thinkingBudget = -1` (Default)
2.5 Flash Lite	Model does not think	512 to 24576	`thinkingBudget = 0`	`thinkingBudget = -1`

3. System Instruction

cURLPython

bash

curl "https://api.exchangetoken.ai/v1beta/models/gemini-3-flash-preview:generateContent" \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "system_instruction": {
      "parts": [
        {
          "text": "You are a cat. Your name is Neko."
        }
      ]
    },
    "contents": [
      {
        "role": "user",
        "parts": [
          {
            "text": "Hello there"
          }
        ]
      }
    ]
  }'

python

import requests
import json

url = "https://api.exchangetoken.ai/v1beta/models/gemini-3-flash-preview:generateContent"
headers = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json"
}

data = {
    "system_instruction": {
        "parts": [
            {"text": "You are a cat. Your name is Neko."}
        ]
    },
    "contents": [
        {
            "role": "user",
            "parts": [
                {"text": "Hello there"}
            ]
        }
    ]
}

response = requests.post(url, headers=headers, json=data)
print(response.json())

4. Multimodal Input

cURLPython

bash

# Use a temporary file to hold the base64 encoded image data
TEMP_B64=$(mktemp)
trap 'rm -f "$TEMP_B64"' EXIT
base64 $B64FLAGS $IMG_PATH > "$TEMP_B64"

# Use a temporary file to hold the JSON payload
TEMP_JSON=$(mktemp)
trap 'rm -f "$TEMP_JSON"' EXIT

cat > "$TEMP_JSON" << EOF
{
  "contents": [
    {
      "role": "user",
      "parts": [
        {
          "text": "Tell me about this instrument"
        },
        {
          "inline_data": {
            "mime_type": "image/jpeg",
            "data": "$(cat "$TEMP_B64")"
          }
        }
      ]
    }
  ]
}
EOF

curl "https://api.exchangetoken.ai/v1beta/models/gemini-3-flash-preview:generateContent" \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -X POST \
  -d "@$TEMP_JSON"

python

import requests
import base64
import json

url = "https://api.exchangetoken.ai/v1beta/models/gemini-3-flash-preview:generateContent"
headers = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json"
}

# Read and encode the image
image_path = "path/to/image.jpeg"
with open(image_path, "rb") as image_file:
    image_data = base64.b64encode(image_file.read()).decode('utf-8')

data = {
    "contents": [
        {
            "role": "user",
            "parts": [
                {"text": "Tell me about this instrument"},
                {
                    "inline_data": {
                        "mime_type": "image/jpeg",
                        "data": image_data
                    }
                }
            ]
        }
    ]
}

response = requests.post(url, headers=headers, json=data)
print(response.json())

Supported Image Formats

Gemini supports the following image format MIME types:

PNG - image/png
JPEG - image/jpeg
WEBP - image/webp
HEIC - image/heic
HEIF - image/heif

To learn about other file input methods, see the guide on file input methods.

5. Streaming Response

cURLPython

bash

curl "https://api.exchangetoken.ai/v1beta/models/gemini-3-flash-preview:streamGenerateContent?alt=sse" \
  -H "Authorization: Bearer $YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  --no-buffer \
  -d '{
    "contents": [
      {
        "role": "user",
        "parts": [
          {
            "text": "Explain how AI works"
          }
        ]
      }
    ]
  }'

python

import requests
import json

# Note: Streaming requests use streamGenerateContent and the alt=sse parameter
url = "https://api.exchangetoken.ai/v1beta/models/gemini-3-flash-preview:streamGenerateContent?alt=sse"
headers = {
    "Authorization": "Bearer YOUR_API_KEY",
    "Content-Type": "application/json"
}

data = {
    "contents": [
        {
            "role": "user",
            "parts": [
                {"text": "Explain how AI works"}
            ]
        }
    ]
}

response = requests.post(url, headers=headers, json=data, stream=True)

for line in response.iter_lines():
    if line:
        print(line.decode('utf-8'))

Gemini (Google Protocol) ​

1. Overview ​

Supported Google Models ​

2. Request Instructions ​

3. Request Parameters ​

3.1 Header Parameters ​

3.2 Body Parameters (application/json) ​

Complete Code Examples ​

1. Text Generation ​

2. Thinking Model ​

Controlling Thinking ​

Thinking Level (Gemini 3) ​

Thinking Budget ​

3. System Instruction ​

4. Multimodal Input ​

Supported Image Formats ​

5. Streaming Response ​

Gemini (Google Protocol)

1. Overview

Supported Google Models

2. Request Instructions

3. Request Parameters

3.1 Header Parameters

3.2 Body Parameters (application/json)

Complete Code Examples

1. Text Generation

2. Thinking Model

Controlling Thinking

Thinking Level (Gemini 3)

Thinking Budget

3. System Instruction

4. Multimodal Input

Supported Image Formats

5. Streaming Response