NeoneAI

NeoneAI is an independent AI research project focused on building efficient language models and exploring new neural network architectures.

Our primary goal is to develop AI systems that can run directly on mobile devices while maintaining strong reasoning abilities and high efficiency. We believe that advanced artificial intelligence should not be limited to massive cloud infrastructure. Powerful AI should also be available on everyday hardware, making intelligent systems more accessible to everyone.

---

Current Models

ARM (Advanced Reasoning Model)

ARM is NeoneAI's core language model lineup, designed for real-world efficiency and reasoning.

Current models:

• ARM-126 — gen-1 126M parameters, 16 layers, width 1024 MLP 1024
• ARM-294 Orion 294M parameters, 24 layers, width 1024 MLP 2048

ARM models are built to:

• Run locally on consumer hardware
• Deliver low-latency responses
• Support instruction tuning and SFT
• Work as edge AI assistants
• Provide transparent and controllable behavior

Training data includes:

• DCLM-Baseline
• FineWeb-Edu
• UltraChat
• SmolTalk
• Custom identity and safety SFT

ARM is the practical foundation of NeoneAI.

---

ALM (Advanced Logic Model)

ALM is the first generation of language models developed by NeoneAI and serves as the technological foundation of the project.

The purpose of ALM is not only to generate natural language, but also to provide a flexible platform for future research in efficient neural architectures. Every new version of ALM is designed with the long-term vision of combining practical performance with significantly lower hardware requirements.

Current research includes:

• Large Language Model training
• Efficient inference
• Instruction tuning
• Reasoning improvements
• Mobile AI optimization
• Long-context research
• Efficient memory utilization
• Scalable model architecture

ALM is being developed through continuous experimentation and iterative improvements. Each training run provides valuable insights that help shape future generations of the model.

---

LNN (Logic Neural Network)

LNN is NeoneAI's long-term research project and represents our vision for the future of artificial intelligence.

While ALM is focused on building highly efficient language models, LNN aims to explore entirely new approaches to neural intelligence, logical reasoning, planning and decision making.

The project investigates ideas that may extend beyond conventional transformer-based architectures. Rather than simply increasing model size, the goal is to improve the quality of reasoning, internal world modelling and computational efficiency.

LNN is currently in its research phase and will continue evolving alongside ALM. Many of the technologies developed during ALM research will contribute to the future development of LNN.

---

Our Vision

We want to build AI that is:

• Efficient
• Practical
• Accessible
• Optimized for mobile devices
• Research-driven
• Designed for long-term development

NeoneAI is a long-term research project. Every milestone is another step toward building more capable, more efficient and more accessible artificial intelligence.

As the project grows, new models, research papers, technologies and development updates will be published here.