Kimi-VL, the open-source multimodal model is redefining what efficient vision-language models can do. At its core is a Mixture-of-Experts (MoE) architecture that activates just 2.8B parameters, yet performs well beyond its weight. Whether it’s cracking complex math problems, parsing ultra-high-resolution images, inferring poorly written manuscripts or engaging in multi-turn agent interactions, Kimi-VL excels across the variety of tasks. With native-resolution visual perception via MoonViT and support for an extended 128K context window, this model processes long, dense, and diverse inputs with clarity and depth. From OCR and video understanding to reasoning across multiple images and documents, Kimi-VL proves itself as a generalist with specialist-level precision. It is competing with top-tier models like GPT-4o-mini and Qwen2.5-VL-7B, and even surpassing them in several specialized tasks while maintaining a lean computational footprint.
And if you’re looking for next-level reasoning, the Kimi-VL-Thinking variant pushes the boundaries even further with long-horizon chain-of-thought capabilities. In this article, we’ll explain the step-by-step process to successfully download and run this model for inference.
Performance
Benchmarks
| Benchmark (Metric) | GPT-4o | GPT-4o-mini | Qwen2.5-VL-72B | Qwen2.5-VL-7B | Gemma-3-27B | Gemma-3-12B | o1-1217 | QVQ-72B | Kimi-k1.5 | Kimi-VL-Thinking-A3B |
|---|
| Thinking Model? | | | | | | | ✅ | ✅ | ✅ | ✅ |
| MathVision (full) (Pass@1) | 30.4 | – | 38.1 | 25.1 | 35.5 | 32.1 | – | 35.9 | 38.6 | 36.8 |
| MathVista (mini) (Pass@1) | 63.8 | 56.7 | 74.8 | 68.2 | 62.3 | 56.4 | 71.0 | 71.4 | 74.9 | 71.3 |
| MMMU (val) (Pass@1) | 69.1 | 60.0 | 74.8 | 58.6 | 64.8 | 59.6 | 77.3 | 70.3 | 70.0 | 61.7 |
Prerequisites
The minimum system requirements for running this model are:
- GPU: RTX A6000
- Storage: 50GB (preferable)
- Jupyter Notebook installed.
- VRAM: 48GB
Step-by-step process to install and run Kimi-VL A3B Thinking
For the purpose of this tutorial, we’ll use a GPU-powered Virtual Machine by NodeShift since it provides high compute Virtual Machines at a very affordable cost on a scale that meets GDPR, SOC2, and ISO27001 requirements. Also, it offers an intuitive and user-friendly interface, making it easier for beginners to get started with Cloud deployments. However, feel free to use any cloud provider of your choice and follow the same steps for the rest of the tutorial.
Step 1: Setting up a NodeShift Account
Visit app.nodeshift.com and create an account by filling in basic details, or continue signing up with your Google/GitHub account.
If you already have an account, login straight to your dashboard.
Step 2: Create a GPU Node
After accessing your account, you should see a dashboard (see image), now:
- Navigate to the menu on the left side.
- Click on the GPU Nodes option.
- Click on Start to start creating your very first GPU node.
These GPU nodes are GPU-powered virtual machines by NodeShift. These nodes are highly customizable and let you control different environmental configurations for GPUs ranging from H100s to A100s, CPUs, RAM, and storage, according to your needs.
Step 3: Selecting configuration for GPU (model, region, storage)
- For this tutorial, we’ll be using 1x RTX A6000 GPU, however, you can choose any GPU as per the prerequisites.
- Similarly, we’ll opt for 200GB storage by sliding the bar. You can also select the region where you want your GPU to reside from the available ones.
Step 4: Choose GPU Configuration and Authentication method
- After selecting your required configuration options, you’ll see the available GPU nodes in your region and according to (or very close to) your configuration. In our case, we’ll choose a 1x RTX A6000 48GB GPU node with 64vCPUs/63GB RAM/200GB SSD.
2. Next, you’ll need to select an authentication method. Two methods are available: Password and SSH Key. We recommend using SSH keys, as they are a more secure option. To create one, head over to our official documentation.
Step 5: Choose an Image
The final step is to choose an image for the VM, which in our case is Jupyter, where we’ll deploy and run the inference of our model using Diffusers.
That’s it! You are now ready to deploy the node. Finalize the configuration summary, and if it looks good, click Create to deploy the node.
Step 6: Connect to active Compute Node using SSH
- As soon as you create the node, it will be deployed in a few seconds or a minute. Once deployed, you will see a status Running in green, meaning that our Compute node is ready to use!
- Once your GPU shows this status, navigate to the three dots on the right, click on Connect with SSH, and copy the SSH details that appear.
Step 7: Setting up Python Notebook
Start by creating a .ipynb notebook by clicking on Python 3 (ipykernel).
Next, If you want to check the GPU details, run the following command in the Jupyter Notebook cell:
!nvidia-smi
Step 8: Download and Set Up Model Dependencies
- Install Python dependencies to run the model.
!pip install torch torchvision torchaudio einops timm pillow
!pip install transformers==4.48.2
!pip install git+https://github.com/huggingface/accelerate
!pip install git+https://github.com/huggingface/diffusers
!pip install huggingface_hub
!pip install sentencepiece bitsandbytes protobuf decord numpy
Output:
2. Install project dependencies.
!pip install tiktoken blobfile
Output:
Step 9: Download and Run the Model
- Finally, we’ll download the model files and load them for inference with the code snippet below:
from PIL import Image
from transformers import AutoModelForCausalLM, AutoProcessor
model_path = "moonshotai/Kimi-VL-A3B-Thinking"
model = AutoModelForCausalLM.from_pretrained(
model_path,
torch_dtype="auto",
device_map="auto",
trust_remote_code=True,
)
processor = AutoProcessor.from_pretrained(model_path, trust_remote_code=True)
Output:
2. Run the model with prompt images for inference.
image_paths = ["./demo1.png", "./demo2.png"]
images = [Image.open(path) for path in image_paths]
messages = [
{
"role": "user",
"content": [
{"type": "image", "image": image_path} for image_path in image_paths
] + [{"type": "text", "text": "Please infer step by step who this manuscript belongs to and what it records"}],
},
]
text = processor.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt")
inputs = processor(images=images, text=text, return_tensors="pt", padding=True, truncation=True).to(model.device)
generated_ids = model.generate(**inputs, max_new_tokens=2048)
generated_ids_trimmed = [
out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids)
]
response = processor.batch_decode(
generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False
)[0]
print(response)
In the above code, we have asked the model to observe, detect and infer what’s written in the poorly written manuscripts as shown below:
Once it successfully runs, you’ll be able to see the generated response as shown below:
Conclusion
Kimi-VL and its advanced variant, Kimi-VL-Thinking, showcase how far efficient multimodal models have come, delivering powerful reasoning, high-resolution visual understanding, and long-context processing with minimal compute. These capabilities open the door to a wide range of real-world applications, from research and education to automation and digital agents. Running such cutting-edge models becomes seamless with NodeShift Cloud, which offers GPU-optimized infrastructure and a developer-friendly environment to deploy, test, and scale models like Kimi-VL effortlessly.
