Expansive language models face a challenge: their sheer size. Multiverse Computing, an emerging Spanish firm, tackles this predicament using condensed models, striving to bridge the divide between the capabilities of cutting-edge models and what enterprises can realistically finance for implementation.
At the heart of their innovation lies CompactifAI, a data compaction technique drawing inspiration from quantum computing, which this Basque enterprise has employed on models originally launched by OpenAI. Currently, programmers are able to freely obtain an enhanced iteration of Multiverse’s HyperNova 60B model via Hugging Face. Furthermore, the firm intends to publicly release additional compact models in 2026, facilitating a broader spectrum of applications.
As per Multiverse, their models possess a reduced footprint, yet maintain comparable power and precision. Weighing in at 32GB, HyperNova 60B measures approximately fifty percent of the volume of its originating model — OpenAI’s gpt-oss-120B — simultaneously offering diminished memory consumption and quicker response times. The refreshed variant, known as HyperNova 60B 2602, presently offers enhanced compatibility with tool invocation and autonomous coding, scenarios where inferencing expenses can escalate.
Among the rivals Multiverse asserts to have surpassed with HyperNova 60B is Mistral Large 3, a model originating from the French decacorn Mistral AI. Nevertheless, beyond this technical competition, the pair of European AI enterprises share numerous similarities.
Similar to Mistral, Multiverse has broadened its presence past its native land, establishing branches in the United States, Canada, and throughout Europe. Both organizations similarly cater to corporate clients. For Multiverse, notable clients include Iberdrola, Bosch, and the Bank of Canada.
Although Multiverse has not yet attained official unicorn status, reports now circulate that it is seeking to secure a new €500 million investment round, potentially valuing the firm at over €1.5 billion. In an announcement provided to TechCrunch, the enterprise verified that active negotiations with prospective backers for a new funding injection are progressing, but stated that discussing valuation or the capital amount at this juncture would be premature. Multiverse moreover chose not to address claims that its yearly recurring revenue (ARR) reached €100 million in January.
Should this be validated, it would nonetheless represent merely a segment of OpenAI’s $20 billion ARR; yet, it wouldn’t be vastly dissimilar from Mistral’s, whose ARR climbed past $400 million, partly due to increasing appetite for non-U.S. technological options. Correspondingly, Multiverse presents itself in its most recent media statement as an entity capable of furnishing “independent solutions across the AI infrastructure.”
Techcrunch event
Boston, MA
|
June 9, 2026
Such geopolitical nuances recently aided Multiverse in solidifying a partnership with the regional administration of Aragón, located in Spain’s northeastern sector. The Spanish Agency for Technological Transformation (SETT) similarly took part in the AI startup’s $215 million Series B funding round during the previous year. From its foundation, Multiverse has furthermore received aid from the Basque territory — which might soon celebrate its inaugural unicorn.
{content}
Source: {feed_title}