Market impact includes enabling SMEs, edge deployments, regulated industries, and research institutions to access advanced AI, potentially accelerating adoption across multiple sectors.

Multiverse is reportedly pursuing a new funding round around €500 million that could value the company above €1.5 billion, though formal ARR figures remain unconfirmed.

At about 32GB, HyperNova 60B matches or closely approaches the accuracy of the full model, with improved tool-calling and agentic coding features now accessible on Hugging Face.

Multiverse has a global footprint with offices in Europe, North America, and partnerships with Iberdrola, Bosch, and the Bank of Canada, plus collaborations with Aragón’s regional government.

Multiverse Computing released HyperNova 60B 2602, a 50% compressed version of GPT-OSS-120B, freely available on Hugging Face to boost efficiency with minimal performance loss.

Industry experts see compression as essential for sustainable AI, with the compression market projected to grow significantly as efficient models gain prominence.

The 60B-2602 model retains near-identical tool-calling capabilities to the larger 120B while operating at half the size, reducing memory from 61GB to 32GB.

Documentation, benchmarks, and integration guides are published on Hugging Face to help IT leaders evaluate performance, security, and fit before large-scale rollouts.

The 2602 upgrade enhances tool calling and agentic coding for high-cost workloads that strain compute budgets.

The company positions itself around sovereign AI and an expanding open-source model release strategy to reduce dependence on closed stacks and support enterprise experimentation.

The release aligns with Europe’s sovereign AI push, emphasizing efficiency, open access, and local experimentation to reduce dependence on U.S.-based providers.

Enterprise-friendly, locally runnable models are highlighted to preserve data residency and governance, contrasting with reliance on external hyperscale services.