Multiverse Computing Accelerates Mainstream Adoption with Compressed AI Models

What’s New: CompactifAI and Quantum‑Inspired Compression

Multiverse Computing, a Spanish AI startup, has unveiled a suite of compressed AI models built with its proprietary CompactifAI technology. The company’s flagship mobile app, also called CompactifAI, demonstrates how a model the size of a few megabytes—named Gilda—can run entirely on‑device, delivering instant responses without sending data to a remote server.

The app automatically switches to a cloud‑based fallback (e.g., the OpenAI ChatGPT integration) when the device lacks sufficient RAM or storage, ensuring a smooth user experience while preserving privacy whenever possible.

In addition to the consumer‑facing app, Multiverse launched a self‑serve API portal that gives developers direct access to the compressed models, removing the need for third‑party marketplaces.

Why Enterprises Care: Privacy, Edge‑vs‑Cloud, and Cost Savings

For enterprise AI teams, the value proposition of Multiverse’s approach is threefold:

• Data privacy: On‑device inference means sensitive information never leaves the corporate network, a critical advantage for regulated sectors such as finance and healthcare.
• Edge resilience: Devices operating in remote locations—drones, satellites, or factory floor robots—can continue to function even when connectivity is intermittent.
• Reduced compute spend: Smaller models consume less GPU/CPU time, translating into lower cloud‑billing and a smaller carbon footprint.

The fallback mechanism, codenamed Ash Nazg, intelligently routes requests to a cloud model when local resources are insufficient, preserving functionality without sacrificing the privacy‑first promise.

Multiverse’s HyperNova 60B 2602 model, derived from the open‑source gpt‑oss‑120b, exemplifies how compression can retain high‑quality output while cutting inference latency by up to 40% compared with the original model.

Latest Model Releases & Real‑World Use Cases

Since the launch of CompactifAI, Multiverse has released three notable compressed models:

1. Gilda 7B: A 7‑billion‑parameter model that fits on most modern smartphones, ideal for chat assistants and quick summarization tasks.
2. HyperNova 30B: Optimized for code generation and reasoning, it powers AI‑assisted development tools in low‑latency environments.
3. HyperNova 60B 2602: The most powerful compressed offering, used for complex analytics and multi‑turn conversational agents.

Enterprise deployments already showcase the technology’s versatility:

• Bank of Canada: Uses on‑device fraud detection models to analyze transaction patterns without transmitting raw data.
• Bosch: Embeds compressed vision‑language models in factory robots, enabling real‑time defect detection even in offline mode.
• Iberdrola: Deploys edge AI for predictive maintenance of wind turbines, where network coverage is spotty.

These examples illustrate how AI on device can unlock new business models—think autonomous drones that process video streams locally, or field‑service tablets that provide instant expert advice without a data‑center connection.

Funding, Customers, and Competitive Landscape

Multiverse Computing raised a $215 million Series B in 2025, led by Lux Capital and other European VCs. Industry insiders report that the company is now courting a €500 million round at a valuation north of €1.5 billion, positioning it as one of the most valuable AI compression startups in Europe.

With a customer base exceeding 100 global enterprises, Multiverse is rapidly moving from a niche “model‑compression” play to a full‑stack Enterprise AI platform contender. Its focus on privacy‑first edge inference differentiates it from larger LLM providers that still rely heavily on massive data‑center deployments.

Competitors such as Mistral AI and Meta are also releasing smaller models, but Multiverse’s unique quantum‑inspired compression pipeline and the UBOS partner program for integration partners give it a strategic edge in the fast‑growing edge AI market.

How Your Business Can Leverage Edge AI Today

If you’re a tech decision‑maker looking to experiment with on‑device models, UBOS offers a ready‑made ecosystem that can accelerate your time‑to‑value:

• Explore the UBOS templates for quick start, including the “AI SEO Analyzer” and “AI Article Copywriter” that run locally on your workstation.
• Build custom workflows with the Workflow automation studio, integrating compressed models via the Chroma DB integration for vector search.
• Deploy voice‑enabled assistants using the ElevenLabs AI voice integration alongside on‑device inference.
• Leverage the Web app editor on UBOS to prototype a compact AI chatbot, similar to the “GPT‑Powered Telegram Bot” template.
• Scale your solution with the UBOS pricing plans that include enterprise‑grade support and SLA guarantees.

Whether you’re a startup (UBOS for startups) or an SMB (UBOS solutions for SMBs), the platform’s modular architecture lets you plug in Multiverse‑style compressed models via the Telegram integration on UBOS or the ChatGPT and Telegram integration.

Ready to bring AI to the edge? Visit the UBOS homepage and start building today.

Source & Further Reading

The details above are based on the original report from TechCrunch. For a deeper dive into edge‑AI trends, explore UBOS’s Edge AI blog series.