Instructions to use SandLogicTechnologies/Internvl3-8b-instruct-GGUF with libraries, inference providers, notebooks, and local apps. Follow these links to get started.

Libraries

How to use SandLogicTechnologies/Internvl3-8b-instruct-GGUF with llama-cpp-python:

# !pip install llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
	repo_id="SandLogicTechnologies/Internvl3-8b-instruct-GGUF",
	filename="InternVL3-8B-Instruct_IQ3_S.gguf",
)

llm.create_chat_completion(
	messages = [
		{
			"role": "user",
			"content": "What is the capital of France?"
		}
	]
)

Notebooks
Google Colab
Kaggle
Local Apps

llama.cpp

How to use SandLogicTechnologies/Internvl3-8b-instruct-GGUF with llama.cpp:

Install from brew

brew install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S
# Run inference directly in the terminal:
llama-cli -hf SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S

Install from WinGet (Windows)

winget install llama.cpp
# Start a local OpenAI-compatible server with a web UI:
llama-server -hf SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S
# Run inference directly in the terminal:
llama-cli -hf SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S

Use pre-built binary

# Download pre-built binary from:
# https://github.com/ggerganov/llama.cpp/releases
# Start a local OpenAI-compatible server with a web UI:
./llama-server -hf SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S
# Run inference directly in the terminal:
./llama-cli -hf SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S

Build from source code

git clone https://github.com/ggerganov/llama.cpp.git
cd llama.cpp
cmake -B build
cmake --build build -j --target llama-server llama-cli
# Start a local OpenAI-compatible server with a web UI:
./build/bin/llama-server -hf SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S
# Run inference directly in the terminal:
./build/bin/llama-cli -hf SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S

Use Docker

docker model run hf.co/SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S

LM Studio
Jan

vLLM

How to use SandLogicTechnologies/Internvl3-8b-instruct-GGUF with vLLM:

Install from pip and serve model

# Install vLLM from pip:
pip install vllm
# Start the vLLM server:
vllm serve "SandLogicTechnologies/Internvl3-8b-instruct-GGUF"
# Call the server using curl (OpenAI-compatible API):
curl -X POST "http://localhost:8000/v1/chat/completions" \
	-H "Content-Type: application/json" \
	--data '{
		"model": "SandLogicTechnologies/Internvl3-8b-instruct-GGUF",
		"messages": [
			{
				"role": "user",
				"content": "What is the capital of France?"
			}
		]
	}'

Use Docker

docker model run hf.co/SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S

Ollama
How to use SandLogicTechnologies/Internvl3-8b-instruct-GGUF with Ollama:
```
ollama run hf.co/SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S
```

Unsloth Studio new

How to use SandLogicTechnologies/Internvl3-8b-instruct-GGUF with Unsloth Studio:

Install Unsloth Studio (macOS, Linux, WSL)

curl -fsSL https://unsloth.ai/install.sh | sh
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for SandLogicTechnologies/Internvl3-8b-instruct-GGUF to start chatting

Install Unsloth Studio (Windows)

irm https://unsloth.ai/install.ps1 | iex
# Run unsloth studio
unsloth studio -H 0.0.0.0 -p 8888
# Then open http://localhost:8888 in your browser
# Search for SandLogicTechnologies/Internvl3-8b-instruct-GGUF to start chatting

Using HuggingFace Spaces for Unsloth

# No setup required
# Open https://huggingface.co/spaces/unsloth/studio in your browser
# Search for SandLogicTechnologies/Internvl3-8b-instruct-GGUF to start chatting

Pi new

How to use SandLogicTechnologies/Internvl3-8b-instruct-GGUF with Pi:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S

Configure the model in Pi

# Install Pi:
npm install -g @mariozechner/pi-coding-agent
# Add to ~/.pi/agent/models.json:
{
  "providers": {
    "llama-cpp": {
      "baseUrl": "http://localhost:8080/v1",
      "api": "openai-completions",
      "apiKey": "none",
      "models": [
        {
          "id": "SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S"
        }
      ]
    }
  }
}

Run Pi

# Start Pi in your project directory:
pi

Hermes Agent new

How to use SandLogicTechnologies/Internvl3-8b-instruct-GGUF with Hermes Agent:

Start the llama.cpp server

# Install llama.cpp:
brew install llama.cpp
# Start a local OpenAI-compatible server:
llama-server -hf SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S

Configure Hermes

# Install Hermes:
curl -fsSL https://hermes-agent.nousresearch.com/install.sh | bash
hermes setup
# Point Hermes at the local server:
hermes config set model.provider custom
hermes config set model.base_url http://127.0.0.1:8080/v1
hermes config set model.default SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S

Run Hermes

hermes

Docker Model Runner
How to use SandLogicTechnologies/Internvl3-8b-instruct-GGUF with Docker Model Runner:
```
docker model run hf.co/SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S
```

Lemonade

How to use SandLogicTechnologies/Internvl3-8b-instruct-GGUF with Lemonade:

Pull the model

# Download Lemonade from https://lemonade-server.ai/
lemonade pull SandLogicTechnologies/Internvl3-8b-instruct-GGUF:IQ3_S

Run and chat with the model

lemonade run user.Internvl3-8b-instruct-GGUF-IQ3_S

List all available models

lemonade list

InternVL3-8B-Instruct

InternVL3-8B-Instruct is an advanced multimodal large language model (MLLM) developed by OpenGVLab, designed for vision-language reasoning, conversational AI, coding workflows, document understanding, OCR, and structured multimodal analysis. This repository contains GGUF quantized variants of the model optimized for efficient local inference using llama.cpp.

The model combines a vision encoder with a large language model through native multimodal pre-training, enabling strong image understanding and multimodal reasoning capabilities while maintaining competitive language performance. Compared to previous InternVL generations, InternVL3 improves multimodal perception, reasoning consistency, long-context capability, and structured visual understanding.

Model Overview

Model Name: InternVL3-8B-Instruct
Base Model: OpenGVLab/InternVL3-8B-Instruct
Architecture: Vision Encoder + Decoder-only Transformer
Parameter Count: ~8B parameters
Context Window: Extended multimodal context support
Modalities: Text, Image
Primary Languages: English, Chinese, Multilingual
Developer: OpenGVLab
License: MIT

Quantization Formats

This repository provides various GGUF quantized versions of the InternVL3-8B-Instruct model, optimized for efficient local inference using llama.cpp. Below are the details of the available I-Matrix (IQ) formats.

IQ3_S

Size reduction of approx 77.03% (3.26 GB) compared to 16-bit (14.19 GB)
Lightweight 3-bit quantization optimized for reducing memory usage while retaining practical multimodal reasoning capability
Suitable for edge deployments, constrained GPU environments, and CPU-based inference workflows
Enables efficient execution of multimodal reasoning and OCR-oriented tasks on lower-memory hardware
Complex visual reasoning, dense document analysis, and long multimodal interactions may experience reduced output fidelity due to aggressive compression

IQ4_NL

Size reduction of approx 70.75% (4.15 GB) compared to 16-bit (14.19 GB)
Advanced 4-bit non-linear quantization designed to better preserve multimodal reasoning quality and structured outputs
More suitable for document understanding, OCR workflows, coding assistance, and analytical image reasoning tasks
Typically provides stronger consistency and better visual-text alignment compared to lower-bit formats
Slightly increased computational overhead during inference due to more sophisticated reconstruction methods

IQ4_XS

Size reduction of approx 72.09% (3.96 GB) compared to 16-bit (14.19 GB)
Balanced 4-bit quantization focused on efficient inference and stable multimodal generation performance
Provides a practical trade-off between memory efficiency, response quality, and inference speed
Suitable for conversational AI, image understanding, OCR, and structured multimodal reasoning workflows
Maintains reliable performance across most real-world multimodal and vision-language workloads

Training Background (Original Model)

InternVL3-8B-Instruct is trained using a native multimodal pre-training strategy that jointly learns linguistic and multimodal capabilities from large-scale multimodal and text-only datasets.

Pretraining

Large-scale multimodal and text-only pretraining
Joint optimization of language and visual understanding capabilities
Focus on image reasoning, OCR, document understanding, and contextual multimodal perception
Optimized for conversational, analytical, and multimodal reasoning workloads

Instruction Tuning

Refined using instruction-following and multimodal reasoning datasets
Enhanced for structured visual reasoning and conversational workflows
Improved consistency for coding, OCR, GUI reasoning, and document analysis tasks
Further optimized using multimodal preference optimization strategies

Key Capabilities

Multimodal Understanding Supports combined text and image reasoning for vision-language tasks and conversational workflows.
Document and OCR Understanding Performs effectively on document parsing, OCR-related reasoning, and structured visual-text analysis.
Reasoning and Analysis Handles multi-step analytical reasoning across both textual and visual inputs.
Coding and Technical Assistance Supports code explanation, structured reasoning, and technical problem-solving workflows.
Long-Context Multimodal Processing Maintains contextual consistency across extended multimodal interactions and large visual-text inputs.
Efficient Local Deployment Quantized GGUF variants enable practical local multimodal inference on consumer hardware.

Usage Example

Using llama.cpp

./llama-cli \
  -m SandlogicTechnologies/InternVL3-8B-Instruct_IQ4_NL.gguf \
  -p "Analyze the uploaded document image and summarize the key findings."

Recommended Usecases

Multimodal AI Assistants Build conversational systems capable of handling both text and image inputs.
Document Understanding and OCR Process scanned documents, screenshots, tables, and structured visual content.
Visual Reasoning Workflows Perform analytical reasoning across charts, diagrams, and technical images.
Coding and Technical Assistance Support structured coding workflows and technical reasoning tasks.
Research and Experimentation Evaluate multimodal prompting strategies and local vision-language inference pipelines.

Acknowledgments

These quantized models are based on the original work by the OpenGVLab development team.

Special thanks to:

The OpenGVLab team for developing and releasing the InternVL3-8B-Instruct model.
Georgi Gerganov and the llama.cpp open-source community for enabling efficient quantization and inference via the GGUF format.

Contact

For questions, feedback, or support, please reach out at support@sandlogic.com or visit https://www.sandlogic.com/

Downloads last month: 581

GGUF

Model size

8B params

Architecture

qwen2

Hardware compatibility

3-bit

4-bit

Model tree for SandLogicTechnologies/Internvl3-8b-instruct-GGUF

Base model

OpenGVLab/InternVL3-8B-Pretrained

Finetuned

OpenGVLab/InternVL3-8B-Instruct

Quantized

(5)

this model