Multiverse Computing Launches Compressed AI Model to Compete in the Market

Multiverse Computing, a Spanish startup, is addressing the challenges associated with large language models by introducing compressed AI models. The company has developed a compression technology named CompactifAI, inspired by quantum computing, which it has applied to models released by OpenAI. The latest iteration, the HyperNova 60B model, is now available for free on Hugging Face, allowing developers to access a more efficient alternative to traditional models.

TechCrunch

The HyperNova 60B model is approximately half the size of OpenAI’s gpt-oss-120b model, at 32GB, while maintaining comparable potency and accuracy. This reduction in size translates to lower memory usage and latency, making it more accessible for companies looking to deploy AI solutions. The updated version, HyperNova 60B 2602, enhances support for tool calling and agentic coding, which are critical for applications with high inference costs.

TechCrunch

Multiverse Computing positions itself as a competitor to other AI firms, notably Mistral AI, which has released the Mistral Large 3 model. Both companies have expanded their operations internationally, with Multiverse establishing offices in the United States, Canada, and various European countries. Notable enterprise clients for Multiverse include Iberdrola, Bosch, and the Bank of Canada. While Multiverse is not yet a unicorn, it is reportedly in discussions to raise €500 million at a valuation exceeding €1.5 billion.

TechCrunch

Although Multiverse's annual recurring revenue (ARR) is estimated at €100 million, it remains significantly lower than OpenAI's $20 billion ARR and Mistral's $400 million ARR. However, the company aims to carve out a niche by offering "sovereign solutions across the AI stack," appealing to clients seeking alternatives to U.S.-based technology.

TechCrunch

Multiverse Computing has confirmed ongoing discussions with potential investors regarding a new funding round but refrained from commenting on the specifics of the valuation or funding size. The company has also received support from the Basque region and secured a collaboration with the regional government of Aragón, which underscores the strategic importance of its innovations in the context of local technological advancement.

TechCrunch

While Multiverse Computing is gaining traction, it faces skepticism regarding its ability to compete with established players like OpenAI and Mistral AI. Critics may question whether the compressed models can truly match the performance of larger counterparts, particularly in high-stakes applications.

TechCrunch

• “According to Multiverse, its models are smaller, but nearly as potent and accurate.” — Multiverse Computing Representative

TechCrunch

Looking ahead, Multiverse plans to open source additional compressed models in 2026, aiming to broaden its applicability across various use cases and further establish its presence in the AI market.

TechCrunch