Multiverse Computing Launches Compressed AI Models for Mass Adoption

After compressing models from major AI labs including OpenAI, Meta, DeepSeek, and Mistral AI, Multiverse Computing has launched an app and API portal to showcase and make its compressed models more widely available. The Spanish startup, which has kept a lower profile, is now targeting businesses with a self-serve API portal that gives developers direct access to its compressed models. The CompactifAI app, featuring a small model called Gilda, is an AI chat tool that runs locally and offline. However, it is limited by the user's device's RAM and storage, and switching to cloud-based models via API is necessary when the user’s device does not meet the required specifications. The company claims that its latest compressed model, HyperNova 60B 2602, is faster and more cost-effective than the original model it was derived from, particularly for agentic coding workflows.

Related coverage

Continue reading with more reporting from the same topic cluster.

• Two Palantir veterans just came out of stealth with $30 million and a Sequoia stamp of ap…
• Meta is having trouble with rogue AI agents
• Sam Altman’s thank-you to coders draws the memes