Spanish AI startup Multiverse Computing has unveiled a free, compressed version of its large language model, HyperNova 60B 2602, available on Hugging Face. The move aims to make powerful AI more accessible and affordable for developers and enterprises by tackling one of the biggest hurdles in AI deployment: model size. The company’s breakthrough lies in CompactifAI, a compression technology inspired by principles from quantum computing. Applied to models originally developed by OpenAI, including the gpt-oss-120b, CompactifAI reduces the model footprint without sacrificing performance. The new HyperNova 60B 2602 version weighs in at just 32GB—about half the size of its source model—while delivering comparable accuracy, lower memory usage, and reduced latency. The updated model also enhances support for tool calling and agentic coding, areas where inference costs typically spike. Multiverse claims its model outperforms Mistral Large 3, a top-tier offering from French decacorn Mistral AI, particularly in efficiency and responsiveness. Beyond technical performance, Multiverse and Mistral share striking similarities. Both are European AI leaders expanding rapidly beyond their home markets, with offices in the U.S., Canada, and across Europe. They serve major enterprise clients: Multiverse counts Iberdrola, Bosch, and the Bank of Canada among its partners. While Multiverse is not yet officially a unicorn, it is reportedly raising a fresh €500 million funding round at a valuation exceeding €1.5 billion. The company confirmed ongoing investor discussions but declined to comment on specific figures or valuation details. It also did not confirm reports that its annual recurring revenue (ARR) reached €100 million in January—though if true, that would place it well behind OpenAI’s $20 billion ARR but in the same ballpark as Mistral’s over $400 million. Multiverse positions itself as a provider of “sovereign solutions across the AI stack,” emphasizing data privacy, regional control, and compliance—key concerns for governments and institutions. This strategic angle helped secure a collaboration with the regional government of Aragón in northeastern Spain, and the Spanish Agency for Technological Transformation (SETT) was a key backer in Multiverse’s $215 million Series B last year. With roots in the Basque region and strong regional support, Multiverse is emerging as a leading candidate to become Spain’s first AI unicorn. The company plans to open source additional compressed models in 2026, broadening access to high-performance AI for developers worldwide.