Multiverse Computing opens CompactifAI API to developers
Get the Tech newsletter
Daily tech — startups, AI labs, chips, the launches that shape the next decade. Free.
- Multiverse Computing launched a self-serve API portal giving developers and enterprises direct access to its compressed AI models without going through AWS Marketplace, according to CEO Enrique Lizaso.
- The CompactifAI app features "Gilda," a model the company says can run locally and offline on mobile devices, though devices without sufficient RAM or storage are routed to cloud models via a system Multiverse has named "Ash Nazg" — and that fallback erases the privacy advantage.
- Multiverse has compressed models from OpenAI, Meta, DeepSeek, and Mistral AI, with its latest release HyperNova 60B 2602 built on OpenAI's open-source gpt-oss-120b and claiming faster responses at lower cost than the original.
- The CompactifAI app had fewer than 5,000 downloads in the past month according to Sensor Tower data, a sign that mass consumer adoption isn't the immediate target.
- Multiverse already serves more than 100 global customers including Bank of Canada, Bosch, and Iberdrola, and is reportedly raising a €500 million round at a €1.5 billion-plus valuation after closing a $215 million Series B last year.
- The launch lands as private company defaults hit 9.2% — the highest in years — prompting VC firm Lux Capital to warn AI companies to get compute capacity commitments confirmed in writing rather than relying on handshake deals.
Why it matters: On-device AI removes the counterparty risk that comes with cloud-dependent inference, an angle that gains weight as private company defaults climb to a multi-year high of 9.2%. Multiverse's 100-plus enterprise roster and rumored €500M raise at €1.5B-plus signal that compressed, edge-deployable models are drawing serious capital — not just consumer hype.
