Large language models face a significant challenge: their sheer size. Multiverse Computing, a startup based in Spain, is tackling this problem by developing compressed models designed to bridge the divide between the capabilities of cutting-edge models and what businesses can practically implement within budget. The key innovation is CompactifAI, a compression technique influenced by quantum computing principles, which this Basque firm has used to optimize models originally released by OpenAI.

Starting today, developers can freely access an updated iteration of Multiverse’s HyperNova 60B model on Hugging Face. The company has further announced its intention to open-source additional compressed models in 2026, aiming to cater to a broader spectrum of applications. Multiverse asserts that its models are substantially smaller yet retain nearly equivalent power and accuracy. For instance, the HyperNova 60B model occupies about 32GB, approximately half the size of its progenitor, OpenAI’s gpt-oss-120B model, while also offering reduced memory consumption and lower latency.

The latest version, designated HyperNova 60B 2602, now provides enhanced support for tool calling and agentic coding—areas where inference expenses can be considerable. Multiverse states it has outperformed one of its competitors, Mistral Large 3, a model from the French decacorn Mistral AI, with its HyperNova 60B. Beyond this technological competition, however, the two European AI firms share notable similarities.

Similar to Mistral, Multiverse has established an international presence with offices in the United States, Canada, and various European locations. Both companies also serve enterprise clients; Multiverse counts Iberdrola, Bosch, and the Bank of Canada among its customers. Although Multiverse has not yet officially attained unicorn status, reports suggest it is currently seeking a new funding round of approximately €500 million, which would value the company at over €1.5 billion.

The company has chosen not to comment on claims that its annual recurring revenue hit €100 million in January. If verified, this figure would still represent only a small portion of OpenAI’s $20 billion ARR, yet it places Multiverse relatively close to Mistral, whose ARR reportedly surged past $400 million—growth partly fueled by increasing demand for alternatives to U.S. technology.

In a similar strategic positioning, Multiverse describes itself in recent communications as an organization capable of “delivering sovereign solutions across the AI stack.” These geopolitical themes recently aided the company in securing a partnership with the regional government of Aragón in northeastern Spain. Additionally, the Spanish Agency for Technological Transformation (SETT) contributed to the AI startup’s $215 million Series B funding round last year.

Since its founding, Multiverse has also received consistent support from the Basque region, which may soon celebrate its first homegrown unicorn company.