Name: Gemini 3.1 Pro API: Multimodal Reasoning LLM
Brand: magicapi
SKU: api-magicapi-advanced-reasoning-ai-google-gemini-3-1-pro
Availability: InStock
Author: API Market

Shashank Agarwal

# 🧠 Multimodal Reasoning LLM API (Google Gemini 3.1 Pro)

## Overview
Gemini 3.1 Pro is a large language model from Google. Google describes it as its most intelligent model, with improved reasoning and a new medium thinking level.

Send a text `prompt` and, optionally, up to 10 `images`, up to 10 `videos` and one `audio` file. Choose how deeply the model reasons with `thinking_level` (`low`, `medium` or `high`). The answer comes back as a list of text chunks; join them in order to get the full text.

## Features
- **Thinking levels**: `thinking_level` sets the maximum depth of the model's internal reasoning: `low`, `medium` or `high` (default).
- **Image input**: up to 10 image URLs in `images`, each up to 7MB.
- **Video input**: up to 10 video URLs in `videos`, each up to 45 minutes. `video_fps` (0.1 to 60) sets how many frames per second are sampled; leave it out for the default of 1 fps. More frames give the model more to reason over and use more tokens.
- **Audio input**: one audio file of up to 8.4 hours in `audio`.
- **System instruction**: steer the role, tone or format of the answer with `system_instruction`.
- **Long answers**: `max_output_tokens` goes up to 65535, which is also the default.
- **Sampling controls**: `temperature` (0 to 2, default 1) and `top_p` (0 to 1, default 0.95).

## Use Cases
1. **Code generation and review**: ask for code in a given language, or paste code into the prompt and ask for an explanation or a fix.
2. **Questions about images**: send screenshots, charts, documents or photos with a question about them.
3. **Video and audio understanding**: summarize a recording, find where something happens in a video, or answer questions about a meeting.
4. **Multi-step reasoning**: planning, math and analysis tasks where `high` thinking helps.

## Visual Examples
### Example Input
Prompt: "Write a robust asteroid defense system in Rust", with `thinking_level` `high` and `max_output_tokens` `2048`. No images, videos or audio were attached.

```json
{
  "top_p": 0.95,
  "images": [],
  "prompt": "Write a robust asteroid defense system in Rust",
  "videos": [],
  "temperature": 1,
  "thinking_level": "high",
  "max_output_tokens": 2048
}
```

### Example Output
The model returned 20 text chunks. Joined, they read:

~~~markdown
Writing a "robust" system in Rust implies leveraging its strongest features: **memory safety, fearless concurrency, rich type systems, and exhaustive error handling**. 

Below is a backend architecture for an Asteroid Defense System (A.D.S). It uses a multi-threaded event loop, trait-based dependency injection (for different types of sensors and weapons), and strict error handling.

### The Architecture
1. **Sensors (Radar/Lidar):** Run on separate threads, scanning space and sending telemetry via channels.
2. **Command & Control (C2):** The main event loop. It calculates Time-To-Impact (TTI), classifies threats, and assigns weapon systems.
3. **Weapons Systems (Lasers/Missiles):** Thread-safe actuators that attempt to neutralize threats.

### The Rust Code

```rust
use std::sync::mpsc::{self, Receiver, Sender};
use std::sync::{Arc, Mutex};
use std::thread;
use std::time::Duration;

// --- 1. CORE TYPES & MATH ---

#[derive(Debug, Clone, Copy)]
struct Vec3 {
    x: f64,
    y: f64,
    z: f64,
}

impl Vec3 {
    fn distance_to_origin(&self) -> f64 {
        (self.x.powi(2) + self.y.powi(2) + self.z.powi(2)).sqrt()
    }

    fn speed(&self) -> f64 {
        self.distance_to_origin() // Reusing math for velocity vector magnitude
    }
}

#[derive(Debug, Clone)]
struct Asteroid {
    id: u64,
    position: Vec3,
    velocity: Vec3, // km/s
    mass: f64,      // metric tons
}

impl Asteroid {
    /// Calculates Time To Impact (TTI) assuming Earth is at (0,0,0)
    fn time_to_impact(&self) -> f64 {
        let distance = self.position.distance_to_origin();
        let speed = self.velocity.speed();
        if speed == 0.0 {
            return f64::INFINITY;
        }
        distance /
~~~

The text ends partway through the code listing. This request set `max_output_tokens` to `2048`; for long code answers, keep the default of `65535` or set a higher value.

## Input Parameters
| Field | Type | Required | Default | Allowed values | Description |
| --- | --- | --- | --- | --- | --- |
| `prompt` | string | Yes | none |  | The text prompt to send to the model |
| `images` | array of strings (URLs) | No | `[]` |  | Input image URLs to send with the prompt (max 10 images, each up to 7MB) |
| `videos` | array of strings (URLs) | No | `[]` |  | Input video URLs to send with the prompt (max 10 videos, each up to 45 minutes) |
| `video_fps` | number | No | none | `0.1` to `60` | Frame rate (frames per second) at which input videos are sampled. Higher values give the model more frames to reason over (useful for fast-moving footage) but use more tokens. Leave empty for the default sampling rate of 1 fps. |
| `audio` | string (URL) | No | none |  | Input audio URL to send with the prompt (max 1 audio file, up to 8.4 hours) |
| `system_instruction` | string | No | none |  | System instruction to guide the model's behavior |
| `thinking_level` | string | No | `high` | `low`, `medium`, `high` | Thinking level for reasoning. Controls the maximum depth of the model's internal reasoning. |
| `temperature` | number | No | `1` | `0` to `2` | Sampling temperature between 0 and 2 |
| `top_p` | number | No | `0.95` | `0` to `1` | Nucleus sampling: the model considers the tokens within the top_p probability mass |
| `max_output_tokens` | integer | No | `65535` | `1` to `65535` | Maximum number of tokens to generate |

## Pricing
- **Pay As You Go (PAYG) pricing**: you only pay for what you use.
- **Price**: $0.0023 per API unit.
- **1 API unit = 100 weighted tokens**; an output token counts 1, an input token counts 0.2223.
- That is **$5.1129 per million input tokens** and **$23 per million output tokens**.
- Each request is rounded up to a whole API unit, so the minimum charge is 1 unit ($0.0023).
- Attached images, videos and audio are part of the input token count the model reports.
- **Example cost**: the example request used 8 input tokens and 510 output tokens: 8 x 0.2223 + 510 = 511.78 weighted tokens, rounded up to 6 API units = **$0.0138**.
- **Failed or cancelled predictions are not billed.**
- **GET requests to check prediction status** are free and don't count toward API unit usage.
- **No monthly subscription fee.**

## API Documentation

### Base URL
All API requests go to: `https://prod.api.market/api/v1/magicapi/advanced-reasoning-ai-google-gemini-3-1-pro`

### Authentication
Send your api.market API key in the `x-api-market-key` header:
```
x-api-market-key: YOUR_API_MARKET_KEY
```

### Asynchronous Processing
**IMPORTANT**: This API is asynchronous. The flow is:
1. Send a POST request to `/predictions`. A successful request returns **201 Created** with a prediction `id` and a `status` of `starting` or `processing`.
2. Poll GET `/predictions/{id}` with that ID.
3. Keep polling while `status` is `starting` or `processing`.
4. Stop when `status` is `succeeded` (read `output`), `failed` or `canceled` (read `error`).

A polling interval of 1 second and an overall timeout of 600 seconds (10 minutes) work well.

### Endpoints

#### Create a Prediction
```bash
curl -X POST "https://prod.api.market/api/v1/magicapi/advanced-reasoning-ai-google-gemini-3-1-pro/predictions" \
  -H "Content-Type: application/json" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY" \
  -d '{
    "version": "google/gemini-3.1-pro",
    "input": {
      "top_p": 0.95,
      "images": [],
      "prompt": "Write a robust asteroid defense system in Rust",
      "videos": [],
      "temperature": 1,
      "thinking_level": "high",
      "max_output_tokens": 2048
    }
  }'
```

Response (201 Created):
```json
{
  "id": "018v7m3769rmy0d1c6s9qzapq4",
  "version": "google/gemini-3.1-pro",
  "status": "starting",
  "created_at": "2026-10-22T03:40:44.364Z"
}
```

#### Check Prediction Status
```bash
curl -X GET "https://prod.api.market/api/v1/magicapi/advanced-reasoning-ai-google-gemini-3-1-pro/predictions/{id}" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY"
```

Response when the prediction has finished (200 OK):
```json
{
  "id": "018v7m3769rmy0d1c6s9qzapq4",
  "status": "succeeded",
  "output": [
    "Writing a \"robust\" system in Rust implies",
    " leveraging its strongest features: **memory safety, fearless concurrency, rich type systems, and exhaustive error handling**. \n\nBelow is a",
    " backend architecture for an Asteroid Defense System (A.D.S). It uses a multi-threaded event loop, trait",
    "-based dependency injection (for different types of sensors and weapons), and strict error handling.\n\n### The Architecture\n1. **Sensors ("
  ],
  "error": null,
  "created_at": "2026-10-22T03:40:44.364Z",
  "started_at": "2026-10-22T03:40:44.377Z",
  "completed_at": "2026-10-22T03:41:02.005Z"
}
```

`output` is a list of text chunks. Join them in order to get the full answer. The response above is shortened to the first 4 of the 20 chunks; the joined text is shown under Example Output.

### Status Codes
- **201 Created**: prediction created (POST `/predictions`). This is the normal success response.
- **200 OK**: prediction details returned (GET `/predictions/{id}`).
- **400 Bad Request**: invalid input, for example a missing `prompt`, a `thinking_level` that is not `low`, `medium` or `high`, or more than 10 images.
- **401 Unauthorized**: missing or invalid API key.
- **404 Not Found**: no prediction with that ID.
- **429 Too Many Requests**: rate limit exceeded.
- **500 Internal Server Error** or **503 Service Unavailable**: temporary server problem; retry later.

## Rate Limits
All plans allow 10 requests per second. Requests above that return `429 Too Many Requests`; back off before retrying and honour any `Retry-After` header. Polling a prediction with GET is free but still counts toward the per-second limit, so poll every 1 to 5 seconds rather than in a tight loop.

## Error Handling
- Check the HTTP status code of every request and read the response body for the error message.
- A prediction can also end with `status` set to `failed`; the `error` field then explains why. Failed predictions are not billed.
- Retry 429, 500 and 503 responses with exponential backoff. Fix the input before retrying a 400.

## Implementation Guide
### Example Code

#### JavaScript
```javascript
const apiKey = "YOUR_API_MARKET_KEY";
const baseUrl = "https://prod.api.market/api/v1/magicapi/advanced-reasoning-ai-google-gemini-3-1-pro";

async function createAndWaitForPrediction(input) {
  const response = await fetch(`${baseUrl}/predictions`, {
    method: "POST",
    headers: {
      "Content-Type": "application/json",
      "x-api-market-key": apiKey
    },
    body: JSON.stringify({ version: "google/gemini-3.1-pro", input })
  });

  if (response.status !== 201) {
    throw new Error(`API error: ${response.status} ${await response.text()}`);
  }

  const prediction = await response.json();
  return pollForResults(prediction.id);
}

const delay = (ms) => new Promise((resolve) => setTimeout(resolve, ms));

async function pollForResults(predictionId) {
  const maxTimeout = 600000;
  const pollInterval = 1000;
  const startTime = Date.now();

  while (Date.now() - startTime < maxTimeout) {
    const response = await fetch(`${baseUrl}/predictions/${predictionId}`, {
      headers: { "x-api-market-key": apiKey }
    });

    if (!response.ok) {
      throw new Error(`API error: ${response.status} ${await response.text()}`);
    }

    const prediction = await response.json();

    if (prediction.status === "succeeded") {
      return prediction.output;
    }

    if (prediction.status === "failed" || prediction.status === "canceled") {
      throw new Error(`Prediction ${prediction.status}: ${prediction.error || "Unknown error"}`);
    }

    await delay(pollInterval);
  }

  throw new Error("Prediction timed out after 10 minutes");
}

createAndWaitForPrediction({
    "top_p": 0.95,
    "images": [],
    "prompt": "Write a robust asteroid defense system in Rust",
    "videos": [],
    "temperature": 1,
    "thinking_level": "high",
    "max_output_tokens": 2048
  }).then((output) => console.log(output.join("")));
```

#### Python
```python
import time
import requests

API_KEY = "YOUR_API_MARKET_KEY"
BASE_URL = "https://prod.api.market/api/v1/magicapi/advanced-reasoning-ai-google-gemini-3-1-pro"


def create_and_wait_for_prediction(input_data):
    headers = {"Content-Type": "application/json", "x-api-market-key": API_KEY}
    response = requests.post(
        f"{BASE_URL}/predictions",
        headers=headers,
        json={"version": "google/gemini-3.1-pro", "input": input_data},
    )

    if response.status_code != 201:
        raise Exception(f"API error: {response.status_code} {response.text}")

    return poll_for_results(response.json()["id"])


def poll_for_results(prediction_id):
    max_timeout = 600
    poll_interval = 1
    headers = {"x-api-market-key": API_KEY}
    start_time = time.time()

    while time.time() - start_time < max_timeout:
        response = requests.get(f"{BASE_URL}/predictions/{prediction_id}", headers=headers)

        if response.status_code != 200:
            raise Exception(f"API error: {response.status_code} {response.text}")

        prediction = response.json()

        if prediction["status"] == "succeeded":
            return prediction["output"]

        if prediction["status"] in ("failed", "canceled"):
            raise Exception(f"Prediction {prediction['status']}: {prediction.get('error') or 'Unknown error'}")

        time.sleep(poll_interval)

    raise Exception("Prediction timed out after 10 minutes")


output = create_and_wait_for_prediction({
    "top_p": 0.95,
    "images": [],
    "prompt": "Write a robust asteroid defense system in Rust",
    "videos": [],
    "temperature": 1,
    "thinking_level": "high",
    "max_output_tokens": 2048
})
print("".join(output))
```

## FAQ
- **How do I get the full answer text?**
  - Join the chunks in `output` in order, for example `output.join("")` in JavaScript or `"".join(output)` in Python.
- **Which thinking level should I use?**
  - `high` (the default) gives the deepest reasoning. `medium` and `low` limit how deeply the model reasons before answering.
- **Can I send a video or an audio file?**
  - Yes. Pass up to 10 video URLs in `videos` (each up to 45 minutes) and one audio URL in `audio` (up to 8.4 hours). Raise `video_fps` for fast-moving footage; it increases the input token count.
- **How much does a request cost?**
  - Input tokens cost $5.1129 per million and output tokens $23 per million, billed in units of 100 weighted tokens at $0.0023 each and rounded up per request. The example above cost $0.0138.
- **Why does the example answer stop partway through the code?**
  - The request capped `max_output_tokens` at 2048. Leave it at the default of 65535 or set a higher value for long answers.
- **Is the 201 response from POST an error?**
  - No. 201 Created means the prediction was accepted. Use the returned `id` to poll GET `/predictions/{id}` until `status` is `succeeded`, `failed` or `canceled`.

## Contact Us
If you have any questions, reach out to us through the chat on the bottom right of the page. You can also email our support team at hello@magicapi.com.


Call Google Gemini 3.1 Pro with text, images, video and audio, and pick a low, medium or high thinking level. Pay as you go, no monthly fee.