Text Generation
Transformers
Safetensors
qwen3_moe
Neura Tech AI
Neuron
instruct
llm
transformer
mixture-of-experts
Mixture of Experts
multilingual
24B
Qwen3
Neuron-6x4B-Instruct
conversational
Instructions to use Neura-Tech-AI/Neuron-6x4B-Instruct with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use Neura-Tech-AI/Neuron-6x4B-Instruct with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("text-generation", model="Neura-Tech-AI/Neuron-6x4B-Instruct") messages = [ {"role": "user", "content": "Who are you?"}, ] pipe(messages)# Load model directly from transformers import AutoTokenizer, AutoModelForCausalLM tokenizer = AutoTokenizer.from_pretrained("Neura-Tech-AI/Neuron-6x4B-Instruct") model = AutoModelForCausalLM.from_pretrained("Neura-Tech-AI/Neuron-6x4B-Instruct", device_map="auto") messages = [ {"role": "user", "content": "Who are you?"}, ] inputs = tokenizer.apply_chat_template( messages, add_generation_prompt=True, tokenize=True, return_dict=True, return_tensors="pt", ).to(model.device) outputs = model.generate(**inputs, max_new_tokens=40) print(tokenizer.decode(outputs[0][inputs["input_ids"].shape[-1]:])) - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- vLLM
How to use Neura-Tech-AI/Neuron-6x4B-Instruct with vLLM:
Install from pip and serve model
# Install vLLM from pip: pip install vllm # Start the vLLM server: vllm serve "Neura-Tech-AI/Neuron-6x4B-Instruct" # Call the server using curl (OpenAI-compatible API): curl -X POST "http://localhost:8000/v1/chat/completions" \ -H "Content-Type: application/json" \ --data '{ "model": "Neura-Tech-AI/Neuron-6x4B-Instruct", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'Use Docker
docker model run hf.co/Neura-Tech-AI/Neuron-6x4B-Instruct
- SGLang
How to use Neura-Tech-AI/Neuron-6x4B-Instruct 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 "Neura-Tech-AI/Neuron-6x4B-Instruct" \ --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": "Neura-Tech-AI/Neuron-6x4B-Instruct", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }'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 "Neura-Tech-AI/Neuron-6x4B-Instruct" \ --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": "Neura-Tech-AI/Neuron-6x4B-Instruct", "messages": [ { "role": "user", "content": "What is the capital of France?" } ] }' - Docker Model Runner
How to use Neura-Tech-AI/Neuron-6x4B-Instruct with Docker Model Runner:
docker model run hf.co/Neura-Tech-AI/Neuron-6x4B-Instruct
| library_name: transformers | |
| license: apache-2.0 | |
| base_model: | |
| - Qwen/Qwen3-4B-Instruct-2507 | |
| - Qwen/Qwen3-4B-Thinking-2507 | |
| - Neura-Tech-AI/Neuron-4B-Instruct | |
| pipeline_tag: text-generation | |
| language: | |
| - en | |
| - zh | |
| - hi | |
| - ar | |
| - ja | |
| - ko | |
| - fr | |
| - de | |
| - es | |
| - pt | |
| - it | |
| - ru | |
| - tr | |
| - vi | |
| - th | |
| - id | |
| - ms | |
| - bn | |
| - ur | |
| - ta | |
| - te | |
| - mr | |
| - gu | |
| - pa | |
| - fa | |
| new_version: Neura-Tech-AI/Neuron-6x4B-Instruct | |
| tags: | |
| - Neura Tech AI | |
| - Neuron | |
| - instruct | |
| - llm | |
| - transformer | |
| - mixture-of-experts | |
| - moe | |
| - multilingual | |
| - 24B | |
| - Qwen3 | |
| - Neuron-6x4B-Instruct | |
| - safetensors | |
| - conversational | |
| <div align="center"> | |
| <picture> | |
| <img src="Neuron.png" width="90%" alt="Neuron"> | |
| </picture> | |
| </div> | |
| <hr> | |
| # Neuron-6x4B-Instruct | |
| > A next-generation open-source Mixture of Experts (MoE) language model developed by **Neura Tech AI**. | |
| ## Overview | |
| **Neuron-6x4B-Instruct** is a high-performance instruction-tuned Mixture of Experts (MoE) language model built upon the **Qwen3** architecture family. It features **6 routing experts**, each based on a 4B-scale expert design, providing improved expert specialization, efficient sparse computation, and enhanced multilingual capabilities. | |
| Neuron is designed to deliver strong performance across a wide range of AI workloads while maintaining efficient expert routing for inference. | |
| The model is developed entirely by: | |
| - **Neura Tech AI** | |
| Neuron focuses on delivering a capable multilingual AI assistant with strengths in: | |
| - High-efficiency sparse expert routing | |
| - Advanced logical reasoning | |
| - Software engineering and coding assistance | |
| - Mathematics and scientific problem solving | |
| - Agentic workflows and tool calling | |
| - Long-context document understanding | |
| - Multilingual communication across major world languages | |
| ## Model Architecture & Details | |
| Neuron-6x4B-Instruct utilizes a sparse Mixture of Experts architecture where tokens are dynamically routed to specialized experts during inference, improving efficiency without activating every parameter. | |
| - **Model Name:** Neuron-6x4B-Instruct | |
| - **Base Architecture:** Transformer Decoder (Sparse Mixture of Experts) | |
| - **Parameters:** ~24B Total Parameters | |
| - **Total Experts:** 6 Specialists | |
| - **Expert Size:** ~4 Billion Parameters per Expert | |
| - **Active Parameters:** ~4B Parameters Routed per Token (Dynamic Routing) | |
| - **Context Length:** 262,144 Tokens | |
| - **License:** Apache-2.0 | |
| ## Developer | |
| **Project:** Neuron | |
| **Developed by:** | |
| - Neura Tech AI | |
| ## Base Model Acknowledgment | |
| We sincerely thank the **Qwen Team** for releasing the Qwen3 model family under the Apache 2.0 License, which served as the architectural foundation for this project. | |
| ## Features | |
| - **6-Expert Sparse MoE Architecture:** Dynamic routing across six specialized experts for efficient inference. | |
| - **Instruction Optimized:** Tuned for helpful, structured, and accurate responses. | |
| - **Advanced Reasoning:** Strong performance on multi-step reasoning and analytical tasks. | |
| - **Coding Assistant:** Designed for software development, debugging, scripting, and code generation. | |
| - **Large Context Window:** Supports contexts up to 262K tokens for long documents and conversations. | |
| - **Tool Calling Ready:** Suitable for AI agents, structured outputs, JSON generation, and automation workflows. | |
| - **Multilingual:** Supports a broad range of major global languages. | |
| ## Performance | |
| > Benchmark results will be published after the official evaluation process is completed. | |
| | Category | Status | | |
| |----------|--------| | |
| | Knowledge | Coming Soon | | |
| | Reasoning | Coming Soon | | |
| | Coding | Coming Soon | | |
| | Mathematics | Coming Soon | | |
| | Alignment | Coming Soon | | |
| | Agent | Coming Soon | | |
| | Multilingual | Coming Soon | | |
| ## Intended Use | |
| Neuron-6x4B-Instruct is intended for: | |
| - Conversational AI | |
| - Coding assistants | |
| - AI agents | |
| - Research | |
| - Education | |
| - Content generation | |
| - Translation | |
| - Long-context document analysis | |
| - Software engineering workflows | |
| ## Disclaimer | |
| This model is intended for research and production use where appropriate. Outputs should be reviewed before being relied upon in safety-critical, legal, financial, or medical applications. | |
| --- | |
| # © 2026 Neura Tech AI. All rights reserved. |