Spanish AI startup Multiverse Computing released HyperNova 60B 2602, a free compressed AI model now...
The AMW Read
Updates the player map with a high-efficiency model release and signals the 'efficiency-first' pattern (cross.§B) as a challenger to US-centric scaling dominance.
NoveltySignificance
Foundation Models · Player MapScaling Laws
Spanish AI startup Multiverse Computing released HyperNova 60B 2602, a free compressed AI model now available on Hugging Face. At 32GB and 50% smaller than OpenAI's gpt-oss-120B, it claims to beat Mistral Large 3 with 92% lower memory usage, 2.8× higher throughput, and +65% faster latency while maintaining near-equivalent accuracy. The company, rumored to be raising €500M at a €1.5B+ valuation with €100M ARR, uses quantum-inspired CompactifAI technology to make frontier AI capabilities deployable and affordable for enterprises. This signals a shift from "bigger is better" to efficiency-first AI, where sovereign European alternatives challenge US dominance. 🇪🇺 #AI #LLM #ModelCompression #EuropeanAI
