Spanish Startup Multiverse Computing Launches Free HyperNova 60B 2602 AI Model: 50% Compressed, Outperforms OpenAI and Mistral

Spanish startup Multiverse Computing has launched the free HyperNova 60B 2602 AI model on Hugging Face.

This 60 billion parameter model represents a 50% compression of OpenAI's gpt-oss-120B, slashing size from 61GB to 32GB while preserving elite performance.

Revolutionizing AI with CompactifAI

At its core, CompactifAI employs quantum-inspired math to shrink models up to 95% with just 2-3% precision loss.

HyperNova 60B 2602 boasts dramatic improvements: 5x better agentic tool use on Tau2-Bench and 2x on Terminal Bench Hard compared to prior versions.

It rivals or beats larger models like Mistral Large 3 in tool-calling and coding, with far lower memory and latency.

From Quantum Origins to Global Leader

Founded in Spain's Basque region, Multiverse Computing transitioned from quantum software to AI compression, earning government backing and $215 million in Series B funding.

The company now serves over 100 clients, including Iberdrola, Bosch, and the Bank of Canada.

Empowering Future of Efficient AI

CEO Enrique Lizaso Olmos stated, "The launch of HyperNova 60B 2602 demonstrates compression as an iterative process of improvement, giving developers tools without massive infrastructure."

Future plans include more free open-source models in 2026 for enterprise, edge, and sovereign AI deployments.

This release democratizes high-performance AI, reducing costs, energy use, and barriers to innovation across industries.

Developers can download it immediately via Hugging Face.