Spanish ‘soonicorn’ Multiverse Computing releases free compressed AI model

Multiverse Computing, a Spanish startup positioning itself as Europe's next potential AI unicorn, has released a significantly compressed version of its HyperNova 60B model for free on Hugging Face. The company's proprietary CompactifAI compression technology, inspired by quantum computing principles, enables the model to be roughly half the size (32GB) of the OpenAI GPT-OSS-120B model it's derived from while maintaining competitive performance. This release comes as Multiverse reportedly seeks a €500M funding round at a €1.5B+ valuation, following its $215M Series B last year with participation from Spanish government agencies. The company claims its compressed models help bridge the gap between frontier model capabilities and what enterprises can practically deploy, with customers including Iberdrola, Bosch, and the Bank of Canada.

The updated HyperNova 60B 2602 model specifically enhances support for tool calling and agentic coding workflows—areas where inference costs typically run high due to complex reasoning requirements. Multiverse positions its technology as a sovereign European alternative to U.S. AI dominance, similar to French rival Mistral AI, and claims HyperNova 60B outperforms Mistral Large 3 in certain benchmarks. The company plans to open source more compressed models in 2026 to support broader use cases, leveraging geopolitical demand for non-U.S. AI solutions that recently helped secure a collaboration with Spain's Aragón regional government. With reported annual recurring revenue reaching €100M in January, Multiverse represents a growing European challenge to American AI hegemony while addressing the critical industry problem of model size versus deployment cost.