Multiverse Computing Releases Free Compressed AI Model to Cut Costs and Latency

Large language models are powerful — but expensive to run.

Multiverse Computing, a fast-growing “soonicorn” based in Spain, is tackling that problem with model compression technology designed to make frontier AI more affordable and efficient.

The company’s core innovation, called CompactifAI, uses compression techniques inspired by quantum computing to shrink large models while preserving performance.

What’s New?

Developers can now access an updated version of HyperNova 60B (2602) for free on Hugging Face.

Key improvements include:

• Roughly half the size of its base model

• Lower memory usage

• Reduced latency

• Stronger tool-calling capabilities

• Better support for agentic coding

The model is compressed from OpenAI’s gpt-oss-120b and, according to Multiverse, delivers near-frontier performance at significantly lower compute cost.

The company also plans to open-source more compressed models in 2026.

Competitive Positioning

Multiverse claims its HyperNova model outperforms competitors including Mistral Large 3 from Mistral AI.

Beyond technical rivalry, both companies share similarities:

• Expansion across Europe and North America

• Enterprise customers

• Focus on sovereign AI solutions

Multiverse counts clients such as Iberdrola, Bosch, and the Bank of Canada.

Funding Momentum

Multiverse is reportedly preparing a funding round exceeding €500 million at a valuation above €1.5 billion, though the company says discussions are ongoing.

If confirmed, this would still be smaller than the scale of major AI leaders — but signals strong investor confidence in model efficiency as a competitive advantage.

The company positions itself as delivering “sovereign solutions across the AI stack,” aligning with growing European demand for local AI infrastructure alternatives.

Why It Matters

As AI adoption expands, cost and efficiency are becoming as important as raw model size.

Model compression could be the key to:

• Lower deployment costs

• Faster inference

• Wider enterprise adoption

• Sovereign AI strategies

In the next phase of AI competition, smaller may become smarter.