While the race for giant Large Language Models (LLMs) seems never-ending, a rising star from the Spanish Basque Country, Multiverse Computing, has just disrupted the status quo. By releasing its compressed model, HyperNova 60B, for free on Hugging Face, the company isn’t just offering a new tool, it’s redefining AI efficiency.
Quantum alchemy at the service of compression
The true feat of Multiverse lies in its proprietary CompactifAI technology. Drawing inspiration from quantum computing concepts, this method drastically shrinks model size without sacrificing the “brainpower” behind it.
The results are staggering:
- HyperNova 60B weighs in at just 32 GB.
- It boasts a 50% size reduction compared to its original version (OpenAI’s gpt-oss-120b).
- The Bonus: Performance remains nearly identical, shattering the myth that “bigger” always means “smarter.”
HyperNova 60B: Faster, lighter, cheaper
With the 2602 version, Multiverse is tackling a major pain point for developers: cost and latency. By optimizing memory usage, HyperNova streamlines the deployment of coding agents and the execution of tool-calling functions.
Complex tasks that once required massive server infrastructures can now be handled on more modest hardware, making cutting-edge AI both accessible and economically viable.
Related: Mistral AI: The AI gem reaches 11.7 B € in valuation
A Future “Unicorn” with Giant Ambitions
The technical brilliance is only the tip of the iceberg. Behind the scenes, Multiverse Computing is entering a new financial dimension. The startup is reportedly in the process of raising €500 million.
The Key Figure: This funding round would push its valuation past €1.5 billion, officially granting it “Unicorn” status.
The duel of European champions
While Mistral AI is often cited as the flagship of European AI, Multiverse is positioning itself as a serious contender. Although currently smaller in terms of revenue, the Basque startup holds several strategic advantages:
- Technological Sovereignty: A strong European footprint ensuring data independence.
- Massive State Support: Heavy government backing to turn Southern Europe into a premier tech hub.
- Energy Efficiency: Lower memory usage translates to a smaller carbon footprint—a major selling point for the corporations of tomorrow.
Expert Take: By choosing the open-source path for HyperNova 60B, Multiverse Computing isn’t just being philanthropic; it is ensuring rapid community adoption to become the new standard for compact, high-performance AI.
Source: Techcrunch
