In a remarkable development, a Spanish startup has successfully secured a substantial Series B funding round amounting to €189 million (approximately $215 million). This impressive financial backing is attributed to their groundbreaking technology known as “CompactifAI,” which promises to revolutionize the landscape of artificial intelligence.
CompactifAI is a cutting-edge compression technology inspired by quantum computing principles. It has the potential to shrink the size of large language models (LLMs) by an astonishing 95% while maintaining their performance integrity. This innovation could significantly alter the way AI models are utilized across various industries.
The startup specializes in offering compressed versions of popular open-source LLMs, focusing primarily on smaller models. These include notable names in the AI community, and the company is set to introduce a new version of DeepSeek R1 shortly. They are also actively developing additional open-source and reasoning models, although proprietary models from other companies are not included in their offerings.
These “slim” models, as they are referred to, can be accessed through cloud services or licensed for on-premise deployment. The company claims that their models operate at speeds 4 to 12 times faster than their uncompressed counterparts, resulting in a significant reduction in inference costs ranging from 50% to 80%. For example, the cost of using their Lama 4 Scout Slim model on a cloud platform is notably lower than its full-sized version.
One of the standout features of these models is their compact and energy-efficient design, allowing them to run on a variety of devices, including personal computers, smartphones, vehicles, drones, and even compact systems like Raspberry Pi. This opens up exciting possibilities for innovative applications, such as interactive AI experiences in everyday technology.
The technical expertise behind this startup is impressive. Co-founded by a prominent professor from a renowned physics center in Spain, the company benefits from his extensive research in tensor networks. These computational tools, which simulate quantum computing capabilities on conventional hardware, are increasingly being utilized for deep learning model compression.
The other co-founder, who serves as CEO, brings a wealth of experience from academia and the banking sector, having held significant positions in various financial institutions. This diverse background contributes to the strategic vision of the company.
The Series B funding round was spearheaded by a well-known investment firm, which has a history of supporting successful tech companies. Several other notable investors also participated, reflecting strong confidence in the startup’s potential.
With a portfolio of 160 patents and a growing customer base that includes major corporations, the startup is poised for significant growth. This latest funding round brings their total capital raised to approximately $250 million, positioning them as a formidable player in the AI technology landscape.