In a move set to redefine the future of artificial intelligence, Multiverse Computing — a European AI startup based in Donostia, Spain — has introduced what it calls the world’s smallest yet high-performing AI models. Using its revolutionary CompactifAI technology, the company has proven that bigger doesn’t always mean better in AI.

While most of the AI industry races toward ever-larger large language models (LLMs), Multiverse Computing is taking the opposite approach with quantum-inspired AI compression. Its latest releases — the SuperFly and ChickBrain models — compress massive AI architectures by up to 95% while maintaining or even improving accuracy.

This breakthrough could change how we run on-device AI, edge AI, and AI for IoT — making it faster, cheaper, more private, and available even without internet access.

The Technology Behind the Breakthrough: CompactifAI

The core engine behind this innovation is CompactifAI, a proprietary quantum-inspired model compression method developed by Multiverse Computing’s research team, led by co-founder and Chief Scientific Officer Román Orús.

Unlike traditional compression methods, which often reduce precision or remove data in ways that harm performance, CompactifAI uses quantum tensor network algorithms. These techniques, originally inspired by quantum physics, allow the compression of AI models in a more refined and information-preserving manner.

In tests, CompactifAI has achieved:

  • Up to 95% reduction in model size without significant performance loss.
  • Lower hardware requirements, enabling AI to run on embedded systems and low-power devices.
  • Significant cost savings in both energy consumption and cloud computing.

Orús explains, "We’re creating models that can run locally without constant cloud dependence, maintaining privacy and reducing latency — all while keeping performance at or above the original model’s level."

Meet the Models: SuperFly & ChickBrain

SuperFly

SuperFly is a compressed version of Hugging Face’s SmolLM2 135 model, shrunk from 135 million to just 94 million parameters. Despite its tiny footprint, it can handle voice commands, automate appliances, and run entirely offline — no internet needed.

Imagine telling your washing machine “start quick wash” and having it respond instantly without sending any data to the cloud. That’s the power of on-device AI. SuperFly is perfectly suited for IoT, smart home devices, and industrial edge AI deployments.

ChickBrain

ChickBrain targets more advanced reasoning tasks. Built from Meta’s Llama 3.1 8B model, it’s compressed to 3.2 billion parameters — yet in benchmark tests it outperforms its much larger parent on language comprehension, mathematical reasoning, and general knowledge.

It runs locally on laptops like MacBooks, making it ideal for professionals, developers, and industries that require private AI processing without cloud latency.

Both models are available via AWS API for easy integration into software applications, allowing businesses and developers to embed powerful AI into their products at competitive token costs.

Why This Matters to the AI Industry

The industry is facing growing challenges — data privacy concerns, skyrocketing cloud costs, and the unsustainability of massive LLMs. By shifting from cloud AI to localized AI computation, Multiverse Computing:

  • Improves privacy by keeping data processing on the device.
  • Eliminates latency caused by server round trips.
  • Enables offline AI in remote or secure environments.
  • Lowers operational costs for companies deploying AI at scale.

This isn't just a tech upgrade — this is fundamentally changing AI deployment models for manufacturing, finance, healthcare, consumer electronics, and beyond.

Funding & Strategic Partnerships

In June 2025, Multiverse Computing raised €189 million in Series B funding (around $215M USD), led by Bullhound Capital and joined by HP Tech Ventures, Toshiba, and other investors. This brings their total funding since 2019 to about $250M USD.

The company has formed deep partnerships with Apple, Samsung, Sony, HP, as well as with industry leaders like BASF, Bosch, and Moody’s. These alliances are already paving the way for on-device AI integrations in future smartphones, laptops, and IoT platforms.

About Multiverse Computing

Founded in 2019 by Enrique Lizaso Olmos, Román Orús, Alfonso Rubio, and Samuel Mugel, Multiverse Computing specializes in quantum and quantum-inspired artificial intelligence software. Their solutions are already being deployed in finance, energy, cybersecurity, manufacturing, and sustainability projects, often achieving better energy efficiency and cost savings than traditional AI.

With over 160 patents, a growing team of AI scientists, and an expanding Model Zoo of compressed AI systems, Multiverse Computing is building the foundation for small, powerful, and sustainable AI.

The Future: AI That Fits Anywhere

The launch of SuperFly and ChickBrain marks more than a product release — it represents a paradigm shift in AI. Imagine AI assistants in smartwatches, cars, medical devices, and even sensors — all running offline with enterprise-grade intelligence.

Multiverse Computing’s quantum-inspired AI compression makes that vision possible today. This is AI that’s not just smaller — but smarter, faster, greener, and more secure.