Link Centre - Search Engine and Internet Directory

Helping to share the web since 1996

Multiverse Computing Shrinks AI Models With Quantum-Inspired Compression

Large language models are powerful but notoriously resource intensive. Spanish startup Multiverse Computing is working to tackle this challenge by developing compressed AI models designed to narrow the gap between cutting-edge performance and real-world deployment costs.

a close up of a typewriter with a paper on it

At the core of its approach is CompactifAI, a compression method influenced by quantum-computing principles. The company has applied this technology to models originating from OpenAI. Developers can now freely access an updated release of Multiverse’s HyperNova 60B on the Hugging Face  platform, with additional compressed models expected to be open sourced throughout 2026.

Smaller footprint, similar capability

Multiverse says its compressed models maintain strong accuracy despite reduced size. HyperNova 60B occupies roughly 32GB of memory, about half the footprint of the model it is derived from - gpt-oss-120b - while also offering faster response times and lower memory requirements.

The latest update, HyperNova 60B 2602, improves support for features such as tool calling and agent-style coding workflows, scenarios where inference costs can otherwise become significant.

Competing with European AI players

Multiverse claims HyperNova 60B performs competitively against models like Mistral Large 3 from French AI company Mistral AI. Beyond technical competition, both firms share similar growth trajectories, expanding internationally and building enterprise customer bases.

Multiverse reports clients including Iberdrola, Bosch, and the Bank of Canada. The company now operates offices across Europe and North America.

Funding ambitions and market positioning

Although Multiverse has not yet reached unicorn status, it is reportedly in discussions to raise around €500 million at a valuation exceeding €1.5 billion. In comments shared with TechCrunch, the startup confirmed ongoing talks with investors but declined to disclose valuation or funding details. It also did not comment on reports suggesting annual recurring revenue hit €100 million earlier this year.

Even if verified, that figure would remain modest compared with OpenAI’s estimated $20 billion ARR, yet closer to Mistral AI’s revenue, which has climbed above $400 million amid growing demand for alternatives to U.S. AI providers. Multiverse has similarly emphasised its ability to deliver “sovereign” AI solutions spanning the full technology stack.

Regional support and partnerships

Geopolitical considerations and regional backing have contributed to Multiverse’s momentum. The startup recently secured a collaboration with Spain’s Aragón regional government, while the Spanish Agency for Technological Transformation participated in its $215 million Series B round last year. Support from the Basque region has also played a role in the company’s development - potentially positioning it as the area’s first unicorn in the near future.

Newer Articles

Older Articles

← Back to News Headlines