Introducing the DeepNexus Family: A New Era of AI Sovereignty
Posted by Enosis Labs Team on June 24, 2025
In the fast-paced world of artificial intelligence, we are often forced to choose: the power of massive models or the agility of small systems; the convenience of a closed API or the complexity of an open-source solution. At Enosis Labs, we refuse to accept these compromises. We believe in a future where power is not at odds with control, and speed does not sacrifice precision. After months of intense development and strategic investment, today we mark the beginning of that future.
From 'All In One' to 'All In Control'
This phrase is more than a slogan; it's our foundational thesis. The 'All In One' paradigm of the big AI providers has given us incredible tools, but often at the cost of our autonomy. It delivers a powerful but inscrutable black box. We propose a fundamental shift: 'All In Control'. An ecosystem where developers and businesses are not mere consumers of AI, but architects of their own solutions. It's about reclaiming sovereignty over data, model logic, and the end-user experience. It's about building with confidence, knowing exactly how and why your AI behaves as it does.
To materialize this vision, we knew we couldn't rely on third-party solutions. We had to build from the ground up. Thus the Midnight Family is born.
Our Premium Offering: Power and Performance
These models are the cutting edge of our research and the engine of our enterprise solutions. Trained with an investment exceeding $25,000 USD, they represent our commitment to maximum performance and reliability. Initially, they will be available exclusively on our Moonfall platform, enabling perfect integration and optimization. A public API for broader access will launch soon.
Our Commitment: AI for the Community
Our vision of 'All In Control' would be incomplete without putting powerful tools directly in the hands of the community. That's why we're proud to present the Midnight Mini family. These models are our gift and our bet on a more open, experimental, and decentralized AI ecosystem. They will be free and open source forever, available on Hugging Face in FP16 and GGUF formats for maximum compatibility.
- Midnight Mini High (3B)The embodiment of efficiency. This model is optimized for speed and low consumption, making it perfect for running on consumer hardware. Think of on-device assistants, offline-first applications, or productivity tools that don't depend on a constant cloud connection. It's the AI that lives with you, not in the cloud.
- Midnight Mini High Thinking (4B)Don't let its size fool you. This model is an intellectual giant. Achieving 69% on the MMLU benchmark, it demonstrates reasoning capability that rivals much larger models. Its ability to run on laptops with just 2GB of VRAM democratizes access to high-level AI. It's the perfect tool for students, researchers, and independent developers looking to experiment without barriers. It's proof that power doesn't always require a supercomputer.
Witness the Launch: The Moonfall Premiere
Words can only go so far. To truly see what the Midnight Family is capable of, we invite you to the Project Moonfall presentation premiere. In this streaming event, we'll showcase the models in action, executing complex workflows and demonstrating the synergy between speed, control, and creativity. Don't miss the beginning of this new cycle.