Spanish AI startup Multiverse Computing has released HyperNova 60B 2602, a significantly compressed yet powerful large language model, making it freely available on Hugging Face. This strategic move directly addresses the growing industry challenge of large, resource-intensive models that are costly to deploy. By offering a smaller, more efficient alternative, the company aims to democratize access to high-performance AI for developers and enterprises worldwide.
A Quantum-Inspired Approach to Compression
At the core of this innovation is Multiverse's proprietary CompactifAI technology, which leverages quantum-inspired mathematics to streamline complex neural networks. The process analyzes a model's architecture to identify and preserve only its most information-rich components. This allows for a dramatic reduction in model size without sacrificing core intelligence and reasoning abilities.
Unlike conventional compression techniques that often result in a 20-30% loss in accuracy, CompactifAI maintains precision within a minimal 2-3% margin. This breakthrough enables the creation of models that operate with a fraction of the typical compute, memory, and energy requirements. The technology represents a significant step toward making advanced AI practical for real-world production environments.
Enhanced Capabilities and Performance
The new HyperNova 60B 2602 is a compressed version of OpenAI's gpt-oss-120B, successfully reducing its size by half from 61GB to just 32GB. This latest iteration builds upon its predecessor with substantial improvements in key functionalities. It specifically delivers enhanced performance for tool calling and agentic coding, which are critical for developing sophisticated AI applications.
Performance benchmarks validate these advancements, demonstrating up to a five-fold improvement in agentic tool use and a two-fold increase in agentic coding. These enhancements provide developers with more powerful and efficient tools for creating complex AI agents and automated workflows. The model’s balance of reduced size and heightened capability makes it a compelling alternative for deployment at scale.
Strategic Growth and Market Position
Multiverse Computing is rapidly positioning itself as a key European contender in the global AI landscape, drawing comparisons to French giant Mistral AI. The company already serves an impressive roster of enterprise clients, including industry leaders like Iberdrola, Bosch, and the Bank of Canada. This growing commercial adoption underscores the tangible value and applicability of its efficient AI models.
The startup is also gaining significant financial momentum, with reports of a potential €500 million funding round at a valuation exceeding €1.5 billion. While the company has only confirmed active discussions with investors, such a development would secure its status as a unicorn. This trajectory is bolstered by strong government support, including from Spain's Agency for Technological Transformation.
By releasing powerful models on an open-source basis, Multiverse is championing the concept of AI sovereignty in Europe and beyond. This strategy empowers organizations to build independent AI capabilities, reducing their reliance on a handful of dominant U.S. tech providers. The company’s approach aligns with a broader geopolitical push for greater regional technological autonomy and control.
The launch of HyperNova 60B 2602 marks a pivotal moment in the mission to create more sustainable and accessible artificial intelligence. Through its pioneering compression technology, Multiverse Computing is not only lowering technical and financial barriers but also carving out a vital niche in a highly competitive market. As it continues to innovate, the company is poised to become a cornerstone of Europe’s sovereign AI ecosystem.