Tsinghua University Yang Zhenning Institute for Advanced Study

Fermi Universe has announced FermiQLLM 1.0, calling it the first full-pipeline quantum-enhanced LLM. It applies quantum methods to data representation, model architecture, training, reinforcement learning, and evaluation, built on Qwen, and can train and run on existing GPU infrastructure without a fault-tolerant quantum computer. It is in internal testing.

Internally, it reports inference performance up over 15%, reinforcement learning training cost down over 25%, and combined MATH-500, GPQA-Diamond, BBH scores 10–20% above a comparable open-source base. Founded in May after starting the project in March, Fermi Universe has raised about $14M (RMB 100M) cumulative seed funding at a post-money valuation of about $143M (RMB 1B) and works with the Tsinghua University Yang Zhenning Institute for Advanced Study quantum research team.