
Spanish startup multiverse computing on Thursday he said He collected a huge round of series B in the amount of 189 million euros (about 215 million dollars) by force of technology, which he calls “Compactifai”.
Compactifai is a compression technology inspired by a quantum, which is in a position to reduce LLM size by as much as 95% without affecting the performance of the model.
In particular, Multiverse offers compressed versions of known, open source llm-mainly small models-like Llama 4 Scout, Lama 3.3 70b, Lama 3.1 8b and Mistral Small 3.1. The company plans to soon release the Deepseek R1 version and said that it is working on more open source and reasoning. Restricted OPENAI and others are not supported.
His “slim” models, as the company calls, are available via Amazon Web Services or may have a license to make use of the location. The company claims that its models are from 4x to 12x faster than comparable unpressed versions, which translates into a reduction in the costs of application by 50% to 80%. For example, multiverse says His scout Lama 4 costs 10 cents for a million tokens per AWS in comparison with 14 cents Lama 4 Scout.
Multiverse claims that some of his models could also be so small and energy -saving that they can be launched on computers, telephones, cars, drones, and even your favorite small PC, Raspberry Pi. (Suddenly we imagine these implausible Raspberry Houses Pi Christmas light improved with Santas interactive conversations.
Multiverse has some technical power behind it. He was co -founded by CTO Román Orús, a professor at Donostia International Physics Center in San Sebastián, Spain. Orús is known for his own Pioneering work on tensor networks (Do not confuse with all projects related to AI called TENSOR in Google).
Tensor networks are computing tools that imitate quantum computers, but operate on normal computers. One of their foremost applications is currently the compression of deep learning models.
The second co -founder of Multiverse, its general director Enrique Lizaso Olmos, also has many mathematical degrees and is a university professor. He spent most of his profession in banking and is best generally known as the former Deputy General Director of UNNIM Bank.
Series B was led through Bullhound Capital, which was supported by firms comparable to Spotify, Revolut, DeliveLhero, Avito and Discord. HP Tech Ventures, Settpoint Capital International, CDP Venture Capital, Santander Climate VC, Toshiba and Capital Riesgo de Euskadi – Gruo SPR also participated in the round.
Multiverse claims that he has 160 patents and 100 customers around the world, including Iberdrola, Bosch and Bank of Canada. Thanks to this financing, he has collected around USD 250 million so far.