Xiaomi has officially unveiled the Xuanyan O3, a flagship mobile system-on-chip designed for the company’s highest-end devices.
Xiaomi describes it as an “AI flagship SoC,” with major upgrades in CPU, GPU, memory, and on-device artificial intelligence performance.
According to Xiaomi, the Xuanyan O3 achieved an AnTuTu score of up to 5.22 million points, making it the first flagship SoC claimed to surpass the 5-million-point mark. The chip uses a 10-core CPU design, with multi-core performance exceeding 15,000 points for the first time and representing a claimed 60 percent improvement.
The GPU has also received a major upgrade with Xiaomi’s new G2-Ultra NX graphics processor. Xiaomi says GPU performance has increased by 85 percent while power consumption has been reduced by 64 percent compared with the previous generation.
The Xuanyan O3 is also being positioned as the world’s first mobile processor to support LPDDR6 memory. It offers memory bandwidth of up to 113.8GB/s, representing a 48 percent increase over the Xuanyan O1.
AI performance is another major focus. Xiaomi has completely redesigned the NPU architecture around its MiMo on-device large model, delivering a claimed 200 TOPS of tensor computing power and 3.13 TFLOPS of vector computing power. Xiaomi says overall on-device AI performance has improved by 45 percent.
The company has integrated AI acceleration across multiple parts of the chip rather than relying only on the NPU. The CPU includes dual SME2 acceleration units, while the GPU features eight NX neural-network accelerators. The ISP includes an AI noise-reduction unit, the DPU adds hardware-based AI super-resolution, and the ADSP includes a dedicated AI engine.
These upgrades are intended to improve applications such as on-device AI, hardware-level frame interpolation and night-scene video noise reduction.
Xiaomi has also focused on efficiency and system responsiveness. The Xuanyan O3 uses a unified fusion bus architecture along with advanced metal-routing technology, which Xiaomi says helps achieve a static latency of just 82 nanoseconds.
If this article helped you, please consider supporting our work. Every small contribution keeps Abijita.com independent and running.
The company says the Xuanyan O3 is the result of 459 days of development and represents its latest effort to push flagship mobile processors further toward AI-focused computing.
