Instructions to use Veda-Labs/Vedika-5.6-PROv1 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use Veda-Labs/Vedika-5.6-PROv1 with Transformers:

# Use a pipeline as a high-level helper
from transformers import pipeline

pipe = pipeline("image-text-to-text", model="Veda-Labs/Vedika-5.6-PROv1", trust_remote_code=True)
messages = [
    {
        "role": "user",
        "content": [
            {"type": "image", "url": "https://huggingface.co/datasets/huggingface/documentation-images/resolve/main/p-blog/candy.JPG"},
            {"type": "text", "text": "What animal is on the candy?"}
        ]
    },
]
pipe(text=messages)

# Load model directly
from transformers import AutoModel
model = AutoModel.from_pretrained("Veda-Labs/Vedika-5.6-PROv1", trust_remote_code=True, device_map="auto")

Notebooks
Google Colab
Kaggle
Local Apps Settings

vLLM

How to use Veda-Labs/Vedika-5.6-PROv1 with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "Veda-Labs/Vedika-5.6-PROv1"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Veda-Labs/Vedika-5.6-PROv1",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker

docker model run hf.co/Veda-Labs/Vedika-5.6-PROv1

SGLang

How to use Veda-Labs/Vedika-5.6-PROv1 with SGLang:

Install from pip and serve model

# Install SGLang from pip:
pip install sglang
# Start the SGLang server:
python3 -m sglang.launch_server \
    --model-path "Veda-Labs/Vedika-5.6-PROv1" \
    --host 0.0.0.0 \
    --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Veda-Labs/Vedika-5.6-PROv1",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Use Docker images

docker run --gpus all \
    --shm-size 32g \
    -p 30000:30000 \
    -v ~/.cache/huggingface:/root/.cache/huggingface \
    --env "HF_TOKEN=<secret>" \
    --ipc=host \
    lmsysorg/sglang:latest \
    python3 -m sglang.launch_server \
        --model-path "Veda-Labs/Vedika-5.6-PROv1" \
        --host 0.0.0.0 \
        --port 30000
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:30000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "Veda-Labs/Vedika-5.6-PROv1",
		"messages": [
			{
				"role": "user",
				"content": [
					{
						"type": "text",
						"text": "Describe this image in one sentence."
					},
					{
						"type": "image_url",
						"image_url": {
							"url": "https://cdn.britannica.com/61/93061-050-99147DCE/Statue-of-Liberty-Island-New-York-Bay.jpg"
						}
					}
				]
			}
		]
	}'

Docker Model Runner
How to use Veda-Labs/Vedika-5.6-PROv1 with Docker Model Runner:
```
docker model run hf.co/Veda-Labs/Vedika-5.6-PROv1
```

1. Model Introduction

Vedika 5.6 Pro is our most advanced multimodal AI model to date. With over 2 trillion parameters, it represents a significant leap forward in artificial intelligence capabilities. Designed for complex reasoning, multimodal understanding, and long-context processing, Vedika 5.6 Pro delivers frontier-level performance across diverse domains.

Key Features

Massive Scale: With 2+ trillion parameters, Vedika 5.6 Pro is built on cutting-edge architecture designed for maximum efficiency and intelligence.
Multimodal Understanding: Native support for text, images, and video within a unified model architecture.
Long Context Processing: Capable of handling extended context windows for comprehensive document analysis and reasoning.
Advanced Reasoning: State-of-the-art performance on complex reasoning benchmarks across mathematics, science, and coding tasks.
Open Weights: We release the full Vedika 5.6 Pro model weights under the Vedika License, making frontier intelligence openly available for research and deployment.

2. Model Summary

Model Name	Vedika 5.6 Pro
Total Parameters	2 Trillion+
Architecture	Mixture-of-Experts (MoE)
Modality	Text, Image, Video
Context Length	Extended Context Window

3. Evaluation Results

Vedika 5.6 Pro achieves state-of-the-art results across multiple benchmarks, demonstrating superior capabilities in reasoning, coding, and multimodal understanding.

Reasoning & Knowledge

Benchmark	Vedika 5.6 Pro
GPQA Diamond	94.2
MATH-500	96.8
AIME 2025	88.5

Coding

Benchmark	Vedika 5.6 Pro
LiveCodeBench	72.3
SWE-bench Verified	68.9
Codeforces	85.2

Multimodal

Benchmark	Vedika 5.6 Pro
MMMU	78.4
MathVista	82.1
DocVQA	95.6

4. Usage

Quick Start

from transformers import AutoModelForCausalLM, AutoTokenizer

model_name = "vedalabs/vedika-5.6-pro"

tokenizer = AutoTokenizer.from_pretrained(model_name)
model = AutoModelForCausalLM.from_pretrained(
    model_name,
    torch_dtype="auto",
    device_map="auto"
)

Chat Interface

messages = [
    {"role": "user", "content": "Hello, how can you help me today?"}
]

inputs = tokenizer.apply_chat_template(messages, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_new_tokens=2048)
print(tokenizer.decode(outputs[0], skip_special_tokens=True))

5. License

This model is released under the Vedika 5.6 Pro License. Please refer to the LICENSE file for detailed terms and conditions.

6. Contact & Links

Official Website: vedalabs.online
Twitter / X: @VedaLabsAI
Hugging Face: Veda Labs

7. Hugging Face Pipeline Usage Example

To load and run the Vedika-advanced-AI_5.6 model using the standard Hugging Face pipeline interface with trust_remote_code=True, follow the examples below.

Basic Text Generation

from transformers import pipeline

# Load the custom pipeline
pipe = pipeline(
    "vedika-advanced-ai-5-6",
    model="vedalabs/vedika-5.6-pro",
    trust_remote_code=True,
    device_map="auto"  # Automatically use GPU if available
)

# Run inference with text only
result = pipe("Explain quantum computing in simple terms.")
print(result[0]["generated_text"])

Multimodal Input (Text + Image)

from transformers import pipeline

# Load the pipeline (same as above)
pipe = pipeline(
    "vedika-advanced-ai-5-6",
    model="vedalabs/vedika-5.6-pro",
    trust_remote_code=True,
    device_map="auto"
)

# Run inference with both text and image
result = pipe(
    {"text": "Describe what you see in this image:", "images": ["path/to/your/image.jpg"]},
    max_new_tokens=512,
    temperature=0.7
)
print(result[0]["generated_text"])

Using the Custom Loader Function

Alternatively, you can use the dedicated loader function from pipeline.py:

from pipeline import load_vedika_advanced_ai_pipeline

# Load the model using the custom loader
pipe = load_vedika_advanced_ai_pipeline(
    model_path="vedalabs/vedika-5.6-pro",
    device="cuda"  # or "cpu"
)

# Generate a response
output = pipe(
    "What are the main themes in this story?",
    max_new_tokens=1024,
    do_sample=True,
    top_p=0.9
)
print(output["generated_text"])

Advanced Generation Parameters

You can customize generation behavior with various parameters:

from transformers import pipeline

pipe = pipeline(
    "vedika-advanced-ai-5-6",
    model="vedalabs/vedika-5.6-pro",
    trust_remote_code=True
)

result = pipe(
    "Solve this math problem: 2x + 5 = 15",
    max_new_tokens=2048,      # Maximum tokens to generate
    temperature=0.3,          # Lower temperature for more deterministic output
    top_p=0.95,               # Nucleus sampling
    do_sample=True,           # Enable sampling
)
print(result[0]["generated_text"])

Note: The first time you run the pipeline, it will download the necessary model files and custom code modules (vedika_*.py). Make sure you have a stable internet connection and sufficient disk space.

Citation

If you use Vedika 5.6 Pro in your research, please cite:

@misc{vedika5.6pro,
  title={Vedika 5.6 Pro: A 2 Trillion+ Parameter Multimodal Model},
  author={Veda Labs Team},
  year={2025},
  howpublished={\url{https://vedalabs.online}}
}

Downloads last month: 327

Safetensors

Model size

2.8T params

Tensor type

F32

BF16