On October 28, Iluvatar CoreX hosted its 2026 Media Open Day in Shanghai, where it showcased the first live demonstration of the Tianji 300. This milestone marks the Tianji 300 as China’s inaugural general-purpose GPU to deliver real-world performance on par with the H200. Independent testing confirms that the Tianji 300 achieves 1.05 times the overall throughput of the H200 on a single-card basis. The IXL144 super node, integrating 144 GPUs, stands as the industry’s highest-density single-cabinet GPU super node. Additionally, Iluvatar CoreX successfully deployed China’s first GPU+XPU heterogeneous AFD separation solution, slashing inference latency by 60%. On the software front, Iluvatar CoreX adopts an "AI-for-AI" approach and has become the first domestic GPGPU manufacturer to achieve deep integration with FlyDSL. Furthermore, the Tongyang TY1200 outperforms the Thor T5000 in real-world applications, pioneering GPU-accelerated physical simulation via MuJoCo-Warp on a domestically produced edge device.
