HyperNova 60B leads Artificial Analysis ranking for energy-efficient frontier AI
DONOSTIA-SAN SEBASTIÁN, Spain, June 11, 2026 (GLOBE NEWSWIRE) -- Multiverse Computing's HyperNova 60B (version 2605) has been independently evaluated by Artificial Analysis and ranks as the most parameter-efficient model in the 40B–150B open-weights size class. Across the 15 models in the cohort, HyperNova 60B is the only one to combine an Intelligence Index score above 29 with a parameter count of 60B or below. It is also the only European-origin model to land in what Artificial Analysis terms the "most attractive quadrant" of its intelligence-versus-size scatter.
HyperNova 60B is built by Multiverse Computing using its proprietary CompactifAI compression technology. CompactifAI applies quantum-inspired tensor-network mathematics to reorganise already-trained neural networks, retaining the model's reasoning, instruction-following and tool-use capabilities while removing redundant parameters. HyperNova 60B 2605 is publicly available on Hugging Face under the Apache 2.0 licence.
Open-source AI models like gpt-oss-120B are freely available worldwide, but the question of European sovereignty relies on who optimises and runs them. Multiverse Computing’s AI models, built and patented in Europe, let European businesses do both on their own infrastructure.
The result arrives as European AI policy and enterprise buyers converge on three priorities: lower inference cost, reduced energy and carbon footprint, and on-premise deployment in regulated or sovereignty-sensitive environments. Models in the 100B+ parameter range can require multiple high-end GPUs and a hyperscaler contract to serve in production.
The European Commission's AI gigafactory initiative, the EuroStack proposals and national sovereign-AI procurement rules are pushing enterprises and public bodies to source AI capability that can run on European infrastructure under European governance. Models that require a multi-GPU US-domiciled cloud contract to operate at scale, sit awkwardly against this direction of travel.
HyperNova 60B runs on a single GPU with under 40GB of memory, enabling local deployment in finance, energy, healthcare and public-sector estates where sending data to a US-domiciled cloud is either non-compliant or commercially undesirable.
Enrique Lizaso Olmos, CEO and co-founder, Multiverse Computing
"The market has spent two years assuming that smarter AI has to mean bigger, more energy-intensive models. Artificial Analysis's independent review proves that this assumption is flawed. For businesses around the world, that means they can demand frontier-grade AI on their own infrastructure, incurring a fraction of the energy and hardware costs that many of the largest models require.”
The ranking in detail
Artificial Analysis publishes its Intelligence Index v4.0 in a scatter plot (vertical axis) mapped against total parameter count on a logarithmic scale (horizontal axis). The upper-left "most attractive quadrant" identifies models that deliver high intellectual performance at low parameter cost. HyperNova 60B 2605 is one of only two models to land in that quadrant, and the model with both the lowest parameter count and highest intelligence to do so. Comparable models with similar or higher Intelligence Index scores require between 120B and 222B parameters.
On raw intelligence, HyperNova 60B scores 29.3, placing it 6th in the 40B–150B category, ahead of Mistral Small 4, gpt-oss-120b (low), K2 Think V2, LongCat Flash Lite, Llama 3.3 70B and Llama 4 Scout. On intelligence-per-parameter, no model in the cohort outperforms it.
HyperNova 60B’s Intelligence Index score sits just under 3% below that of gpt-oss-120B at high reasoning effort. This, for the majority of enterprise buyers, presents an acceptable gap in exchange for a model that halves its hardware costs, energy costs and removes the need for hyperscaler contracts. Workloads in finance, energy, manufacturing and the public sector rarely need the absolute frontier of model intelligence. Instead they demand reliable reasoning, on infrastructure the business controls, at a cost that scales. HyperNova 60B is also available through Multiverse Computing’s CompactifAI API. Request access here.
The Intelligence Index v4.0 aggregates ten independent evaluations: GDPval-AA, τ²-Bench Telecom, Terminal-Bench Hard, SciCode, AA-LCR, AA-Omniscience, IFBench, Humanity's Last Exam, GPQA Diamond and CritPt. Selected individual results for HyperNova 60B:
For more about HyperNova 60B visit Multiverse Computing’s website here.
About Multiverse Computing
Multiverse Computing is a leader in sovereign and efficient AI. The company develops fast, efficient, and highly specialized AI models that enable organizations to deploy advanced artificial intelligence securely within their own infrastructure, ensuring full control over data, governance, and compliance. Serving sectors where privacy, reliability, and operational efficiency are critical, Multiverse helps enterprises unlock the value of AI while maintaining sovereignty over their most sensitive information. Headquartered in Donostia-San Sebastián, Spain, with offices in the United States, Canada, and across Europe, Multiverse serves more than 100 global customers, including Iberdrola, Bosch, and the Bank of Canada.
For more information, visit www.multiversecomputing.com
About Artificial Analysis
Artificial Analysis is an independent AI benchmark organisation that evaluates and compares language models across intelligence, performance and price. Its Intelligence Index v4.0, released in late 2025, aggregates ten evaluations spanning agents, coding, scientific reasoning and general knowledge, and was recalibrated so that top-performing models score at or below 50, restoring headroom for measurable improvement. Methodology: artificialanalysis.ai/methodology/intelligence-benchmarking.
Media Contact
Ballou for Multiverse Computing multiversecomputing@balloupr.com
Photos accompanying this announcement are available at:
https://www.globenewswire.com/NewsRoom/AttachmentNg/2946068e-863f-4a64-9ca5-f236afcfc322
https://www.globenewswire.com/NewsRoom/AttachmentNg/a05f209c-4fa4-4da9-a2ff-2ed77be9c90c