Multiverse Computing has announced the release of HyperNova 60B 2602, a 50% compressed variant of OpenAI’s gpt-oss-120B, aimed at enhancing accessibility for developers. This model, made available for free on Hugging Face, builds upon the company’s earlier HyperNova 60B launched in January, with notable improvements in tool calling and agentic coding capabilities. The announcement, made on February 24, 2026, marks Multiverse’s commitment to democratizing high-performance artificial intelligence (AI) for the global developer community.
As developers face constraints from the infrastructure demands of large language models (LLMs), Multiverse is focused on delivering efficient and high-performing models that minimize these barriers. Through its proprietary technology, CompactifAI, the company applies quantum-inspired mathematics to optimize neural networks, allowing for substantial reductions in model size without sacrificing accuracy. The compression technology can reduce model size by as much as 95% while maintaining performance within a 2-3% precision margin, significantly outperforming the industry standard of 20-30% accuracy loss with conventional compression techniques.
“The launch of HyperNova 60B 2602 demonstrates compression as an iterative process of improvement, not a one-time optimization,” said Enrique Lizaso Olmos, CEO of Multiverse Computing. “Each generation of our compressed models pushes the boundaries of what’s possible with efficient AI.” This latest release aims to empower developers to experiment, validate, and deploy efficient AI models without the burden of extensive infrastructure investments.
HyperNova 60B 2602 also reflects enhancements driven by real-world feedback from developers. The model is designed to facilitate advanced workflows and coding applications, showcasing significant improvements in key performance benchmarks. Notably, it features a fivefold enhancement in agentic tool use as measured by the Tau2-Bench, a twofold improvement in agentic coding and terminal use as per the Terminal Bench Hard, and a 1.5x improvement in function calling metrics according to BFCL v4.
This new release indicates that high-level reasoning performance can be achieved with drastically lower memory and energy requirements. HyperNova 60B 2602 retains nearly the same tool-calling capabilities as its full-size counterpart while operating at half the size, reducing from 61GB to 32GB. Such advancements validate the feasibility of deploying compressed AI models in production settings.
Multiverse Computing aims to broaden access to production-ready AI models designed for diverse environments, including enterprise, research, and public sectors. The company has plans to roll out additional open-source models and updates throughout the year, expanding their offering to encompass various sizes and applications, from enterprise systems to edge computing.
Positioned uniquely to deliver comprehensive AI solutions, Multiverse’s open-source strategy supports a growing global community of developers and IT professionals evaluating AI for both internal and commercial purposes. By providing open access to its models, organizations can assess performance, security, and operational compatibility before committing to substantial implementations, facilitating a smoother integration process.
Developers and professionals interested in testing Multiverse’s models can find all releases on Hugging Face at https://huggingface.co/MultiverseComputingCAI. HyperNova 60B 2602 is currently accessible at https://huggingface.co/MultiverseComputingCAI/Hypernova-60B-2602. Each release is accompanied by technical documentation, benchmarks, and integration guides available on the company’s Hugging Face page. More information about compressed AI models can be found at multiversecomputing.com.
About Multiverse Computing
Multiverse Computing is recognized as a pioneer in quantum-inspired AI model compression. The company’s expertise in quantum software led to the creation of CompactifAI, which significantly reduces computing needs and unlocks new AI applications across industries. Based in Donostia, Spain, with offices in the United States, Canada, and Europe, Multiverse serves over 100 global clients, including Iberdrola, Bosch, and the Bank of Canada.
Germany”s National Team Prepares for World Cup Qualifiers with Disco Atmosphere
95% of AI Projects Fail in Companies According to MIT
AI in Food & Beverages Market to Surge from $11.08B to $263.80B by 2032
Satya Nadella Supports OpenAI’s $100B Revenue Goal, Highlights AI Funding Needs
Wall Street Recovers from Early Loss as Nvidia Surges 1.8% Amid Market Volatility
