Name: Qwen3 Reranker 8B API: Search and RAG Result Reranking
Brand: magicapi
SKU: api-magicapi-query-search-results-reranker-tejasunku-qwen3-reranker-8b
Availability: InStock
Author: API Market

Shashank Agarwal

# 🔍 Search Result Reranker API (Alibaba Qwen3 Reranker 8B)

## Overview
Qwen3 Reranker 8B scores how relevant each document is to a query and returns the documents ranked by that score. It runs the Qwen3-Reranker-8B weights from Alibaba's Qwen team; its developer describes it as reranking query search results, with support for instructions.

Send a `query`, a list of `documents` (as a JSON string) and an optional task `instruction`. The model returns a JSON string with each document and its relevance score, highest first, limited to `top_k` results.

## Features
- **Query and document relevance**: every document gets a score for the query; higher means more relevant.
- **Task instructions**: describe the retrieval task in `instruction` (recommended for better results). The default is "Given a web search query, retrieve relevant passages that answer the query".
- **Top results only**: `top_k` (1 to 100, default 5) caps how many ranked documents come back.
- **Batching**: `batch_size` (1 to 32, default 8) sets how many documents are scored together.
- **JSON output**: one JSON string with the instruction, the query, the ranked results and the total number of documents.

## Use Cases
1. **Retrieval-augmented generation (RAG)**: rerank chunks from a vector search before passing the best ones to an LLM.
2. **Site and product search**: reorder keyword or semantic search hits by relevance to the user's query.
3. **Support and FAQ matching**: pick the help articles that best answer a customer question.
4. **Filtering candidates**: drop low-scoring passages before showing or summarizing results.

## Visual Examples
### Example Input
```json
{
  "query": "What is the capital of China",
  "top_k": 5,
  "documents": "[\"The capital of China is Beijing\", \"Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun\"]",
  "batch_size": 8,
  "instruction": "Given a web search query, retrieve relevant passages that answer the query"
}
```

### Example Output
`output` is a JSON string. Parsed, it looks like this:

```json
{
  "instruction": "Given a web search query, retrieve relevant passages that answer the query",
  "query": "What is the capital of China",
  "results": [
    {
      "document": "The capital of China is Beijing",
      "score": 0.457763671875
    },
    {
      "document": "Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun",
      "score": -0.8974609375
    }
  ],
  "total_documents": 2
}
```

## Input Parameters
| Field | Type | Required | Default | Allowed values | Description |
| --- | --- | --- | --- | --- | --- |
| `instruction` | string | No | `Given a web search query, retrieve relevant passages that answer the query` |  | Task instruction for the reranker (recommended for better performance) |
| `query` | string | Yes |  |  | Query text for reranking |
| `documents` | string | No | `["Document 1 text", "Document 2 text", "Document 3 text"]` | JSON array of strings, sent as a string | JSON string containing the list of documents to rerank |
| `top_k` | integer | No | `5` | 1 to 100 | Number of top documents to return |
| `batch_size` | integer | No | `8` | 1 to 32 | Batch size for processing documents |

## Pricing
- **Pay As You Go (PAYG) pricing**: you only pay for what you use.
- **Price**: $0.0069 per API unit.
- **1 API unit = 1 successful request with up to 27 seconds of processing.** A longer job bills one more unit for each started block of 27 seconds (for example, 30 seconds of processing is 2 units, $0.0138).
- Processing time grows with the number and length of `documents`.
- **Example cost**: the example request (2 documents) took 1.2 seconds of processing, which is 1 unit, so $0.0069.
- **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/query-search-results-reranker-tejasunku-qwen3-reranker-8b`

### 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. The first request after an idle period can take several minutes while the model loads.

### Endpoints

#### Create a Prediction
```bash
curl -X POST "https://prod.api.market/api/v1/magicapi/query-search-results-reranker-tejasunku-qwen3-reranker-8b/predictions" \
  -H "Content-Type: application/json" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY" \
  -d '{
    "version": "1fb2f6046d0f044dd6b3a13cc7ff61d33973619716991b25d346b455972cec99",
    "input": {
      "query": "What is the capital of China",
      "top_k": 5,
      "documents": "[\"The capital of China is Beijing\", \"Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun\"]",
      "batch_size": 8,
      "instruction": "Given a web search query, retrieve relevant passages that answer the query"
    }
  }'
```

Response (201 Created):
```json
{
  "id": "m0j5s1af9drg80d1d8vbpkemrm",
  "version": "1fb2f6046d0f044dd6b3a13cc7ff61d33973619716991b25d346b455972cec99",
  "status": "starting",
  "created_at": "2026-10-23T19:21:46.734Z"
}
```

#### Check Prediction Status
```bash
curl -X GET "https://prod.api.market/api/v1/magicapi/query-search-results-reranker-tejasunku-qwen3-reranker-8b/predictions/{id}" \
  -H "x-api-market-key: YOUR_API_MARKET_KEY"
```

Response when the prediction has finished (200 OK):
```json
{
  "id": "m0j5s1af9drg80d1d8vbpkemrm",
  "status": "succeeded",
  "output": "{\n  \"instruction\": \"Given a web search query, retrieve relevant passages that answer the query\",\n  \"query\": \"What is the capital of China\",\n  \"results\": [\n    {\n      \"document\": \"The capital of China is Beijing\",\n      \"score\": 0.457763671875\n    },\n    {\n      \"document\": \"Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun\",\n      \"score\": -0.8974609375\n    }\n  ],\n  \"total_documents\": 2\n}",
  "error": null,
  "created_at": "2026-10-23T19:21:46.734Z",
  "started_at": "2026-10-23T19:21:47.734Z",
  "completed_at": "2026-10-23T19:21:49.734Z"
}
```

`output` is a string containing JSON. Parse it to read `results`, a list of `{document, score}` objects sorted from most to least relevant. Scores are relative and can be negative.

### 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 `query` or a `top_k` above 100.
- **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. Check that `documents` is a valid JSON array encoded as a string. 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/query-search-results-reranker-tejasunku-qwen3-reranker-8b";

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: "1fb2f6046d0f044dd6b3a13cc7ff61d33973619716991b25d346b455972cec99", 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 JSON.parse(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({
  "query": "What is the capital of China",
  "top_k": 5,
  "documents": "[\"The capital of China is Beijing\", \"Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun\"]",
  "batch_size": 8,
  "instruction": "Given a web search query, retrieve relevant passages that answer the query"
}).then((ranked) => console.log(ranked.results));
```

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

API_KEY = "YOUR_API_MARKET_KEY"
BASE_URL = "https://prod.api.market/api/v1/magicapi/query-search-results-reranker-tejasunku-qwen3-reranker-8b"


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": "1fb2f6046d0f044dd6b3a13cc7ff61d33973619716991b25d346b455972cec99", "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 json.loads(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")


ranked = create_and_wait_for_prediction({
    "query": "What is the capital of China",
    "top_k": 5,
    "documents": "[\"The capital of China is Beijing\", \"Gravity is a force that attracts two bodies towards each other. It gives weight to physical objects and is responsible for the movement of planets around the sun\"]",
    "batch_size": 8,
    "instruction": "Given a web search query, retrieve relevant passages that answer the query"
})
print(ranked["results"])
```

## FAQ
- **Why is `documents` a string and not an array?**
  - The field takes a JSON-encoded list. Build the list in your code and send `JSON.stringify(list)` (JavaScript) or `json.dumps(list)` (Python) as the value of `documents`.
- **How do I read the scores?**
  - Higher means more relevant to the query. Results come back sorted from highest to lowest score. Scores are relative and can be negative, as in the example.
- **How many documents can I send?**
  - The schema sets no fixed limit on `documents`; `top_k` (up to 100) limits how many ranked results are returned. Large lists take longer and can add 27-second units.
- **Should I change `instruction`?**
  - Yes, when your task is not web search. Describe what a relevant document looks like for your use case, for example matching support articles to customer questions.
- **What if my prediction fails?**
  - Failed predictions are not billed. Read the `error` field, adjust the input and try again.

## 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.


Rerank search results and RAG passages with Qwen3 Reranker 8B: send a query, documents and an optional instruction, get relevance scores sorted highest first.