On October 18, during the 2026 CCF National Conference on High Performance Computing (CCF HPC2026), Sugon presented ten frontier application results achieved on its Sugon 8000 system. Recreating the evolution of dark matter in the universe, tracking atomic motion inside chip materials, "growing" corn inside a computer, replicating a digital brain: a set of scientific and engineering problems long constrained by computing power is being worked through one by one on this domestically built supercluster, with several of the projects claiming world firsts.

Basic Science: From 4 Trillion Dark Matter Particles to a Trillion Atoms

In basic science, the "Qianyan" cosmological simulation recreated the evolution of the universe using 4 trillion dark matter particles, work that international researchers have described as pushing past the current limits of numerical cosmology.

The MatPL machine learning force field software pushed molecular dynamics simulation at near-quantum-mechanical accuracy to the trillion-atom scale for the first time, setting several world records. A team at Fudan University built the world's first digital twin brain at the scale of 100 billion neurons. Researchers at the University of Science and Technology of China and partner institutions paired the "Jiuzhang" photonic quantum computer with the Sugon 8000 to simulate biomolecules at random phase approximation (RPA) accuracy, and completed scalability tests at more than 5 billion atoms.

Public Services: Weather, Oceans and Crop Breeding

On the public services side, Fengqing Ensemble V1.0, a domestically developed AI global short- to medium-range ensemble forecasting model, is now running operationally. A new integrated weather-climate model, MCV, completes kilometer-scale global numerical forecast calculations in one hour, bringing China's operational capability in both AI-based and numerical weather forecasting to the global front rank.

swLFS has produced the world's first global ocean circulation forecasting system at 1-kilometer resolution. China Agricultural University's "virtual corn" platform has cut breeding cycles from more than a decade to a few years.

Engineering Research: Jet Simulation on 6.77 Trillion Grid Points

In engineering research, a multi-nozzle supersonic jet simulation ran on 6.77 trillion grid points, the largest known simulation of its kind internationally. An AI-assisted wall turbulence simulation achieved high-fidelity solutions for channel turbulence, considered the hardest benchmark case in the field. A deep-earth geophysical exploration simulation fills a gap in engineering-grade applications in China.

One Platform, Validating Domestic Heterogeneous Computing

Delivering all ten results on a single platform demonstrates that a domestically built heterogeneous computing system can handle a wide range of research workloads. As AI moves deeper into research, computing systems have to support model training, data analysis and scientific simulation at the same time, accommodating workflows that combine multiple research methods.

Heterogeneous architectures that split work between CPUs and accelerators have become the standard choice for the world's newest top-tier computing systems. The Sugon 8000 is built on a heterogeneous platform of domestic chips, supporting high-precision scientific computing alongside AI training and inference, which allows scientific, engineering and intelligent computing workloads to be scheduled together. Several of the applications maintained relatively high scaling efficiency at large scale, giving researchers room to push toward broader scope and finer resolution.

That convergence extends into research methodology. MatPL uses machine learning force fields to balance simulation accuracy against scale, AI-assisted fluid computation lowers the cost of calculation, and digital breeding combines AI analysis with biological modeling to screen candidates. As AI takes on a larger share of the workload and new forms of computing become practical, heterogeneous architectures that can keep absorbing new compute units offer a higher ceiling and more headroom.

More Than 300 Applications Deeply Optimized

Putting large-scale computing power to real scientific use requires progress on both system capability and application optimization. As workloads grow, data exchange, coordination efficiency and operational stability all shape the final result. The Sugon 8000 coordinates compute, storage, networking, cooling and software to keep complex research workloads running at scale.

On that foundation, Sugon has worked with research users on code porting, performance tuning and application validation, aligning system capability with specific research requirements. More than 300 priority applications have now been deeply optimized on the Sugon 8000, spanning over 30 research and industrial fields, with more than 20 of them running across the full machine. The ten results announced are representative milestones from that ongoing collaboration.

For a wider pool of innovators, the OneScience platform for developing science foundation models and its scientific computing agents further simplify model development, job configuration and results analysis. Through a core node of the National Supercomputing Internet, scientists, engineers and tech startups can access the Sugon 8000's computing power and application services online, drawing on existing application experience to move more research ideas into computational testing.