Multiverse Computing

🚀 New Open-Weight Release from Multiverse Computing HyperNova 60B is a high-performance, compute-efficient large language model developed by Multiverse Computing. Based on gpt-oss-120b, it delivers top-tier reasoning and generation without top-tier infrastructure costs. ⚡ Why it matters 🔹Efficiency-first optimization 🔹Production-ready for enterprise, research, and public sector 🔹Strong gains across intelligence, speed, and memory efficiency 🧠 When compared to Mistral 3 Large, HyperNova 60B establishes a decisive lead across all key dimensions: 1.9× better reasoning performance +12% higher intelligence index −92% lower memory usage 2.8× higher throughput +65% faster latency This combination of intelligence and efficiency is unprecedented at this scale, making HyperNova 60B the clear leader for organizations that need cutting-edge AI performance without prohibitive infrastructure costs. 🤗 Access the model here: https://lnkd.in/eZ3B66FP

🚀 We are heading to CES Las Vegas 2026! We are excited to announce that Multiverse Computing will be exhibiting at CES 2026 in Las Vegas, NV, from January 6–9. You can find us at Venetian Expo, Hall G, Eureka Park, Booth 60821. This year, we will be showcasing how our compressed AI models make it possible to run state-of-the-art AI with a fraction of the compute, energy, and hardware requirements unlocking real-world impact across industries. ✨ Stop by our booth to discover how Multiverse Computing enables organizations to deploy high-performance AI that is smaller, faster, and more efficient, reducing infrastructure costs while addressing scalability and sustainability challenges. 🎯 Let’s connect at CES! We would love to meet innovators, partners, and industry leaders who are building the next generation of AI. 📍 Venetian Expo, Hall G - Eureka Park, Booth 60821 📆 January 6–9, 2026 #CES2026 #ArtificialIntelligence #AIInnovation #EdgeAI #SustainableAI #MultiverseComputing

🚀 New open-weight release from Multiverse Computing We have released Qwen3-Next-80B-A3B-Thinking-Uncensored, a large-scale reasoning model where political refusal behavior has been removed while preserving safety, performance, and architecture. With this release we are pushing the boundaries of SOTA model editing: 🔹 No architectural changes 🔹 No new parameters or training data 🔹 Identical behavior on non-censored prompts The goal is to give researchers and developers access to powerful reasoning models without hard-coded political refusals, while keeping standard safety constraints intact. We are also releasing the full code and datasets used for evaluating censorship and safety. 🤗 Model: https://lnkd.in/eys_XWma 🗒️ Dataset: https://lnkd.in/eerpzaVN 📊 Evaluation library: https://lnkd.in/eTFASH_K #MultiverseComputing #OpenSource #LLM #AIResearch #Qwen #HuggingFace

Merry Christmas & Happy Holidays from Multiverse Computing! ✨🎄 As the year comes to an end, we would like to thank our team, partners, clients, and the entire community for your trust, collaboration, and support throughout this year. We are proud of what we have built together and excited for what lies ahead. Wishing you a joyful holiday season and a successful, innovative year to come. See you in the new year! 🚀❄️ #HappyHolidays #MerryChristmas #MultiverseComputing

Today we had the pleasure of welcoming Eider Mendoza Larrañaga, General Deputy of Gipuzkoa, and Unai Andueza Iraeta, Deputy of Economic Promotion and Strategic Projects, to our HQ in San Sebastian 🌊 Our leadership team was able to exchange perspectives on innovation, technology, and the role advanced AI can play in strengthening the region’s economic and industrial ecosystem 🏭 We had meaningful conversations around ongoing work at Multiverse Computing, future opportunities, and how public institutions and deep-tech companies can collaborate to drive long-term impact. Beyond the discussions, visits like these are essential for building strong relationships, sharing context, and reinforcing the collaborative spirit that connects institutions, talent, and technology. Thank you, Eider and Unai, for taking the time to visit us in San Sebastian 🤝 Enrique Lizaso Olmos Román Orús Iraia Ibarzabal AMAIA GARMENDIA #MultiverseComputing #AI #Gipuzkoa #SanSebastian

🚀 Multiverse Computing Opens New Office in Madrid! 🇪🇸 We are opening our new Madrid office, a major milestone in our growth journey and a key step in strengthening our presence in one of Europe’s most vibrant tech hubs. 🏙️✨ This strategic space will enable us to collaborate even more closely with spanish partners and clients, accelerate innovation, and bring our CompactifAI compressed AI models to more organizations across sectors. 📍 Located at Paseo de la Castellana 200, our Madrid team of over 60 talented professionals, from software engineering to machine learning and data analytics, will play a central role in advancing efficient and sustainable AI solutions. Leaded by our Madrid site lead Gonzalo Otegui 💡 Since 2019, Multiverse has grown from its roots in San Sebastián to a global footprint with offices in Paris, London, Munich, Milan, Toronto, and San Francisco. This expansion reinforces our mission to deliver ultra-efficient compressed AI models that reduce computational costs and energy consumption without compromising performance. 🔗 Learn more here: https://lnkd.in/enzUqdTa #AI #Madrid #ScaleUp #TechInnovation #EfficientAI #CompactifAI

☄️ Introducing HyperNova 60B by Multiverse Computing We are launching HyperNova 60B a revolutionary high-performance AI model, powered by our CompactifAI technology. In direct benchmarks against Mistral 3 Large, HyperNova 60B delivers outstanding gains: 🧠 +12% higher intelligence index* 💾 -92% lower memory usage ⚡ 2.8× higher throughput 🏎️ +65% faster latency *(average of AIME25, GPQA-D, MMLU-Pro and LiveCodeBench) Making it ideal for enterprises, research institutions, and public sector organizations that demand top-tier AI without top-tier costs. This milestone solidifies Multiverse Computing’s leadership in next-generation AI models marrying intelligence with dramatically reduced infrastructure demands. 🔗 Learn more here: https://lnkd.in/eYGzJfgd #AI #MachineLearning #LLM #TechInnovation #AIatScale #Efficiency #MultiverseComputing

★ Introducing BlackStar 12B by Multiverse Computing At Multiverse Computing, we are redefining what high-performance AI models look like with BlackStar 12B, powered by our CompactifAI technology. In direct benchmarks against Ministral 14B (Mistral AI), BlackStar 12B delivers outstanding gains: 📊 Benchmark highlights +33% higher intelligence index* 54% lower memory usage +47% higher throughput +32% faster latency *(average of AIME25, GPQA-D, MMLU-Pro and LiveCodeBench) These results prove that smaller, smarter, and more efficient models can outperform larger ones unlocking enterprise-grade AI with lower infrastructure costs, faster deployment, and superior real-world performance. BlackStar 12B showcases how Multiverse Computing AI models can lead globally, combining state-of-the-art intelligence with unmatched efficiency. 🔗 Read more about BlackStar 12B and our benchmarks: https://lnkd.in/eYGzJfgd #AI #EfficientAI #LLMs #EnterpriseAI #CompactifAI #MultiverseComputing

Multiverse Computing & EY Partner to Unlock Scalable Agentic AI 🤝 Today we are thrilled to announce a strategic collaboration with EY to enable scalable, efficient, and sustainable agentic AI through breakthrough model compression technology 🌍 By combining EY’s deep expertise in enterprise-scale AI ecosystems with Multiverse Computing’s CompactifAI model compression innovations, we are tackling one of the biggest challenges in autonomous AI: computational cost and sustainability. 👉 This partnership empowers organizations to deploy powerful AI agents that can: ⚡️Operate with significantly reduced compute requirements consuming up to 90% less energy. 📈 Scale across industries more affordably and sustainably. 🎯 Deliver real business impact with practical performance. “Together, we will help organizations implement agentic AI systems that are not only more capable, but also significantly more efficient, affordable, and sustainable.” Rodrigo Hernandez, Global GenAI Director, Multiverse Computing. 💡This collaboration marks a pivotal step toward democratizing advanced AI making autonomous, enterprise-ready AI accessible to organizations of all sizes without compromising performance or cost. Let’s build the future of scalable AI together. 🌐 🔗 Learn more here: https://lnkd.in/edcsMz3A #AI #AgenticAI #ModelCompression #SustainableAI #EnterpriseAI #MultiverseComputing #EY

Multiverse Computing sets the benchmark as Europe’s AI champion 🇪🇺 Our latest breakthrough compressed AI models redefine what’s possible in performance, efficiency, and cost. And we can prove this with data 📊 📈 Outperforming top AI models across EVERY KEY METRIC BlackStar 12B by Multiverse Computing vs Ministral 14B by Mistral AI 🎯Our answers are 33% more accurate. ⚡️We respond 47% faster. 🌱And we do it all while using 33% less energy. Watch the video below and visit here to learn more: https://lnkd.in/eYGzJfgd Smaller. Faster. Cheaper. More capable. By redefining efficiency without compromising intelligence, Multiverse Computing is setting the standard for Europe’s AI future. #AI #EnterpriseAI #CompressedAI #MultiverseComputing #LLMs