You can deploy this model and start earning money today!

unsloth/
Qwen3-VL-4B-Thinking-FP8

Name: Qwen3-VL-4B-Thinking-FP8
Availability: InStock

Qwen/Qwen3-VL-4B-Thinking-FP8

quantized

apache-2.0

fp8

Views

No views yet

[!NOTE] Includes Unsloth chat template fixes!
For llama.cpp, use --jinja

Unsloth Dynamic 2.0 achieves superior accuracy & outperforms other leading quants.

Qwen3-VL-4B-Thinking-FP8

This repository contains an FP8 quantized version of the Qwen3-VL-4B-Thinking model. The quantization method is fine-grained fp8 quantization with block size of 128, and its performance metrics are nearly identical to those of the original BF16 model. Enjoy!

Meet Qwen3-VL — the most powerful vision-language model in the Qwen series to date.

This generation delivers comprehensive upgrades across the board: superior text understanding & generation, deeper visual perception & reasoning, extended context length, enhanced spatial and video dynamics comprehension, and stronger agent interaction capabilities.

Available in Dense and MoE architectures that scale from edge to cloud, with Instruct and reasoning‑enhanced Thinking editions for flexible, on‑demand deployment.

Key Enhancements:

Visual Agent: Operates PC/mobile GUIs—recognizes elements, understands functions, invokes tools, completes tasks.
Visual Coding Boost: Generates Draw.io/HTML/CSS/JS from images/videos.
Advanced Spatial Perception: Judges object positions, viewpoints, and occlusions; provides stronger 2D grounding and enables 3D grounding for spatial reasoning and embodied AI.
Long Context & Video Understanding: Native 256K context, expandable to 1M; handles books and hours-long video with full recall and second-level indexing.
Enhanced Multimodal Reasoning: Excels in STEM/Math—causal analysis and logical, evidence-based answers.
Upgraded Visual Recognition: Broader, higher-quality pretraining is able to “recognize everything”—celebrities, anime, products, landmarks, flora/fauna, etc.
Expanded OCR: Supports 32 languages (up from 19); robust in low light, blur, and tilt; better with rare/ancient characters and jargon; improved long-document structure parsing.
Text Understanding on par with pure LLMs: Seamless text–vision fusion for lossless, unified comprehension.

Model Architecture Updates:

Interleaved-MRoPE: Full‑frequency allocation over time, width, and height via robust positional embeddings, enhancing long‑horizon video reasoning.
DeepStack: Fuses multi‑level ViT features to capture fine‑grained details and sharpen image–text alignment.
Text–Timestamp Alignment: Moves beyond T‑RoPE to precise, timestamp‑grounded event localization for stronger video temporal modeling.

This is the weight repository for Qwen3-VL-4B-Thinking-FP8.

Model Performance

Multimodal performance

Pure text performance

Quickstart

Currently, 🤗 Transformers does not support loading these weights directly. Stay tuned!

We recommend deploying the model using vLLM or SGLang, with example launch commands provided below. For details on the runtime environment and deployment, please refer to this link.

vLLM Inference

Here we provide a code snippet demonstrating how to use vLLM to run inference with Qwen3-VL locally. For more details on efficient deployment with vLLM, please refer to the community deployment guide.

python

# -*- coding: utf-8 -*-
import torch
from qwen_vl_utils import process_vision_info
from transformers import AutoProcessor
from vllm import LLM, SamplingParams

import os
os.environ['VLLM_WORKER_MULTIPROC_METHOD'] = 'spawn'

def prepare_inputs_for_vllm(messages, processor):
    text = processor.apply_chat_template(messages, tokenize=False, add_generation_prompt=True)
    # qwen_vl_utils 0.0.14+ reqired
    image_inputs, video_inputs, video_kwargs = process_vision_info(
        messages,
        image_patch_size=processor.image_processor.patch_size,
        return_video_kwargs=True,
        return_video_metadata=True
    )
    print(f"video_kwargs: {video_kwargs}")

    mm_data = {}
    if image_inputs is not None:
        mm_data['image'] = image_inputs
    if video_inputs is not None:
        mm_data['video'] = video_inputs

    return {
        'prompt': text,
        'multi_modal_data': mm_data,
        'mm_processor_kwargs': video_kwargs
    }


if __name__ == '__main__':
    # messages = [
    #     {
    #         "role": "user",
    #         "content": [
    #             {
    #                 "type": "video",
    #                 "video": "https://qianwen-res.oss-cn-beijing.aliyuncs.com/Qwen2-VL/space_woaudio.mp4",
    #             },
    #             {"type": "text", "text": "这段视频有多长"},
    #         ],
    #     }
    # ]

    messages = [
        {
            "role": "user",
            "content": [
              {
                  "type": "image",
                  "image": "https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3-VL/receipt.png",
              },
              {"type": "text", "text": "Read all the text in the image."},
            ],
        }
    ]

    # TODO: change to your own checkpoint path
    checkpoint_path = "Qwen/Qwen3-VL-4B-Thinking-FP8"
    processor = AutoProcessor.from_pretrained(checkpoint_path)
    inputs = [prepare_inputs_for_vllm(message, processor) for message in [messages]]

    llm = LLM(
        model=checkpoint_path,
        trust_remote_code=True,
        gpu_memory_utilization=0.70,
        enforce_eager=False,
        tensor_parallel_size=torch.cuda.device_count(),
        seed=0
    )

    sampling_params = SamplingParams(
        temperature=0,
        max_tokens=1024,
        top_k=-1,
        stop_token_ids=[],
    )

    for i, input_ in enumerate(inputs):
        print()
        print('=' * 40)
        print(f"Inputs[{i}]: {input_['prompt']=!r}")
    print('\n' + '>' * 40)

    outputs = llm.generate(inputs, sampling_params=sampling_params)
    for i, output in enumerate(outputs):
        generated_text = output.outputs[0].text
        print()
        print('=' * 40)
        print(f"Generated text: {generated_text!r}")

SGLang Inference

Here we provide a code snippet demonstrating how to use SGLang to run inference with Qwen3-VL locally.

python

import time
from PIL import Image
from sglang import Engine
from qwen_vl_utils import process_vision_info
from transformers import AutoProcessor, AutoConfig

if __name__ == "__main__":
    # TODO: change to your own checkpoint path
    checkpoint_path = "Qwen/Qwen3-VL-4B-Thinking-FP8"
    processor = AutoProcessor.from_pretrained(checkpoint_path)

    messages = [
        {
            "role": "user",
            "content": [
              {
                  "type": "image",
                  "image": "https://qianwen-res.oss-accelerate.aliyuncs.com/Qwen3-VL/receipt.png",
              },
              {"type": "text", "text": "Read all the text in the image."},
            ],
        }
    ]

    text = processor.apply_chat_template(
        messages,
        tokenize=False,
        add_generation_prompt=True
    )

    image_inputs, _ = process_vision_info(messages, image_patch_size=processor.image_processor.patch_size)

    llm = Engine(
        model_path=checkpoint_path,
        enable_multimodal=True,
        mem_fraction_static=0.8,
        tp_size=torch.cuda.device_count(),
        attention_backend="fa3"
    )

    start = time.time()
    sampling_params = {"max_new_tokens": 1024}
    response = llm.generate(prompt=text, image_data=image_inputs, sampling_params=sampling_params)
    print(f"Response costs: {time.time() - start:.2f}s")
    print(f"Generated text: {response['text']}")

Generation Hyperparameters

VL

bash

export greedy='false'
export top_p=0.95
export top_k=20
export repetition_penalty=1.0
export presence_penalty=0.0
export temperature=1.0
export out_seq_length=40960

Text