Multiverse Computing Releases HyperNova 60B 2602: 50% Compressed OpenAI Model with Enhanced Tool-Calling — MULTIVERSE COMPUTING, S.L.

Event summary

Multiverse Computing released HyperNova 60B 2602, a 50% compressed version of OpenAI's gpt-oss-120B, on Hugging Face for free.
The model improves tool-calling capabilities by 5x in Tau2-Bench and agentic coding by 2x in Terminal Bench Hard compared to the previous version.
HyperNova 60B 2602 operates at half the size (32GB vs. 61GB) while retaining nearly the same tool-calling capability as OpenAI's full-size model.
Multiverse plans additional open-source model releases throughout 2026, targeting enterprise, research, and public sector deployments.

The big picture

Multiverse Computing is pushing the boundaries of AI model efficiency with its quantum-inspired compression technology, CompactifAI. By releasing HyperNova 60B 2602 for free on Hugging Face, the company is challenging the industry's reliance on massive infrastructure for advanced AI deployment. This move aligns with broader trends toward democratizing AI access and reducing operational costs, particularly as enterprises seek to balance performance with scalability.

What we're watching

Adoption Pace: The pace at which developers and enterprises integrate HyperNova 60B 2602 into production environments will signal the real-world viability of compressed AI models.
Competitive Response: Whether OpenAI or other major AI providers respond with their own compression technologies or partnerships to counter Multiverse's open-source strategy.
Sovereign Solutions: How Multiverse's focus on sovereign AI solutions impacts its positioning in regulated industries and geopolitically sensitive markets.