Qwen3-Reranker-0.6B-seq-cls-fp16-ov

Qwen3-Reranker-0.6B-seq-cls-fp16-ov

• Model creator: tomaarsen
• Original model: Qwen3-Reranker-0.6B-seq-cls

Description

This is Qwen3-Reranker-0.6B-seq-cls model converted to the OpenVINO™ IR (Intermediate Representation) format with weights compressed to FP16.

Compatibility

The provided OpenVINO™ IR model is compatible with:

• OpenVINO version 2025.4.0 and higher
• Optimum Intel 1.26.0 and higher

Running Model Inference with Optimum Intel

1. Install packages required for using Optimum Intel integration with the OpenVINO backend:

pip install "git+https://github.com/huggingface/optimum-intel.git" "torch==2.8" --extra-index-url https://download.pytorch.org/whl/cpu

1. Run model inference:

import torch
from transformers import AutoTokenizer
from optimum.intel import OVModelForCausalLM


def format_instruction(instruction, query, doc):
    prefix = '<|im_start|>system\nJudge whether the Document meets the requirements based on the Query and the Instruct provided. Note that the answer can only be "yes" or "no".<|im_end|>\n<|im_start|>user\n'
    suffix = "<|im_end|>\n<|im_start|>assistant\n<think>\n\n</think>\n\n"
    if instruction is None:
        instruction = (
            "Given a web search query, retrieve relevant passages that answer the query"
        )
    output = f"{prefix}<Instruct>: {instruction}\n<Query>: {query}\n<Document>: {doc}{suffix}"
    return output


model_id = "OpenVINO/Qwen3-Reranker-0.6B-seq-cls-fp16-ov"

model = OVModelForCausalLM.from_pretrained(model_id, use_cache=False, export=False)
tokenizer = AutoTokenizer.from_pretrained(model_id, padding_side="left")
# We recommend enabling flash_attention_2 for better acceleration and memory saving.
# model = AutoModelForSequenceClassification.from_pretrained("tomaarsen/Qwen3-Reranker-0.6B-seq-cls", torch_dtype=torch.float16, attn_implementation="flash_attention_2").cuda().eval()
max_length = 8192

task = "Given a web search query, retrieve relevant passages that answer the query"

queries = [
    "Which planet is known as the Red Planet?",
    "Which planet is known as the Red Planet?",
    "Which planet is known as the Red Planet?",
    "Which planet is known as the Red Planet?",
]

documents = [
    "Venus is often called Earth's twin because of its similar size and proximity.",
    "Mars, known for its reddish appearance, is often referred to as the Red Planet.",
    "Jupiter, the largest planet in our solar system, has a prominent red spot.",
    "Saturn, famous for its rings, is sometimes mistaken for the Red Planet.",
]

pairs = [format_instruction(task, query, doc) for query, doc in zip(queries, documents)]
inputs = tokenizer(
    pairs,
    padding=True,
    truncation=True,
    max_length=max_length,
    return_tensors="pt",
)
logits = model(**inputs).logits.squeeze()
print(logits.tolist())
# [-3.109282970428467, 7.120373725891113, -0.37874650955200195, 3.5416228771209717]

scores = logits.sigmoid()
print(scores.tolist())
# [0.04272596165537834, 0.9991921782493591, 0.406429260969162, 0.9718491435050964]

For more examples and possible optimizations, refer to the Inference with Optimum Intel.

Running Model with OpenAI client and OpenVINO Model Server

1a. Deploy model on Windows using binary package:

mkdir C:\models
ovms.exe --rest_port 8000 --source_model OpenVINO/Qwen3-Reranker-0.6B-seq-cls-fp16-ov --model_repository_path C:\models

1b. Deploy model in a Docker container:

mkdir -p ${HOME}/models
export GPU_ARGS=$(if ls /dev/dri/render* >/dev/null 2>&1; then echo "--device /dev/dri --group-add $(stat -c '%g' /dev/dri/render* | head -n1)"; fi)
docker run -d ${GPU_ARGS} --rm --user $(id -u):$(id -g) -p 8000:8000 -v ${HOME}/models:/models openvino/model_server:latest-gpu --rest_port 8000 --model_repository_path /models --source_model OpenVINO/Qwen3-Reranker-0.6B-seq-cls-fp16-ov

1. Install the client library:

pip install requests "numpy<2"

1. Run the client:

import requests

prefix = "<|im_start|>system\nJudge whether the Document meets the requirements based on the Query and the Instruct provided. Note that the answer can only be \"yes\" or \"no\".<|im_end|>\n<|im_start|>user\n"
suffix = "<|im_end|>\n<|im_start|>assistant\n<think>\n\n</think>\n\n"

instruction = "Given a web search query, retrieve relevant passages that answer the query"
query = "What are the key benefits of machine learning in business?"

documents = [
    "Machine learning benefits businesses with better decisions, task automation, customer personalization, and cost savings.",
    "Quantum computing uses quantum bits for fast calculations in cryptography and simulations.",
]

query_formatted = f"{prefix}<Instruct>: {instruction}\n<Query>: {query}\n"
doc_formatted = [f"<Document>: {doc}{suffix}" for doc in documents]

response = requests.post(
    "http://localhost:8000/v1/rerank",
    json={
        "model": "OpenVINO/Qwen3-Reranker-0.6B-seq-cls-fp16-ov",
        "query": query_formatted,
        "documents": doc_formatted,
    }
)

print(response.json())

Limitations

Check the original model card for limitations.

Legal information

The original model is distributed under Apache License Version 2.0 license. More details can be found in Qwen3-Reranker-0.6B-seq-cls.

Disclaimer

Intel is committed to respecting human rights and avoiding causing or contributing to adverse impacts on human rights. See Intel’s Global Human Rights Principles. Intel’s products and software are intended only to be used in applications that do not cause or contribute to adverse impacts on human rights.