Youshu Quantum, a company established with support from Jiaotong University Shanghai, launched the desktop heterogeneous computing platform UnitarySpark and opened UnitaryLab 2.5 for public beta testing during an event in Shanghai on September 20, 2026, according to QbitAI coverage published on September 27.
The official English name of the company in industry profiles such as Caplight and PitchBook is Youshu Quantum, while the product trademarks are UnitarySpark and UnitaryLab.
The core concept revolves around local quantum scientific computation driven by natural language, where sensitive experiment parameters and default model configurations are retained on the workstation rather than being sent to a public cloud.
UnitarySpark integrates quantum simulation, GPU acceleration, a built-in AI agent, and connection to a real QPU within a single desktop workstation. The goal of this solution is to replace disparate simulator stacks and cloud transitions for modeling tasks related to thermal, electromagnetic, molecular, and financial processes that require simultaneous use of classical and quantum resources.
A unified API for circuits was described, allowing algorithms configured in simulation to transition to partner hardware without needing to rewrite the entire toolset.
Strategic Collaboration
The coverage mentioned strategic partnerships with Shanghai QuantumCTek (Guodun), Shanghai Bose Yuanpei Quantum, and Zhongqi Wuliang for software adaptation, connectivity assurance, and scenario validation.
UnitaryLab 2.5 places an Agent between the user and the computing stack. Researchers can describe a differential equation, quantum algorithm, or engineering simulation in natural language; the Agent then translates this request into mathematical formulations, parameters, and executable tasks, subsequently planning local computations and returning interpreted results.
According to an NVIDIA best practice note, workflows that previously required about a dozen manual coding steps can now be reduced to a few natural language prompts. At the heart of UnitaryLab is the team's Schrödingerization framework, which transforms specific problems involving non-unitary differential equations into unitary evolutions that the quantum system can execute. This method, developed at SJTU, is highlighted as the company's primary algorithmic core, rather than a race for qubit count.
The practical application of UnitaryLab on NVIDIA DGX Spark is cited as an NVIDIA success story, and the company presents the launch as the completion of the 'algorithm plus platform plus base' stack for desktop deployment. For buyers of Chinese quantum tools, the specific news is the availability of the local-priority desktop platform UnitarySpark and the beta version of the NL Agent UnitaryLab 2.5—a ready product, not just a funding headline.