Chat about technology with Xiaomi and change life together.
Xiaomi's MiMo-V2-Flash, a 309B MoE model with hybrid attention and Multi‑Token Prediction acceleration, delivers top‑2 global agent benchmark scores, up to 2× faster inference, and only 2.5% of the cost of comparable closed‑source models, while being fully open‑source.
The article explains Xiaomi's "Hotspot Compilation Acceleration" in Surge OS 3, detailing how the compiler identifies high‑frequency code paths, partitions hot and cold sections, and iteratively optimizes binaries, resulting in a 4% CPU load reduction and a 10% efficiency gain for smoother mobile experiences.
A paper co‑authored by Xiaomi’s mobile RF team, Nengxun High‑Energy Semiconductor and HKUST was selected for IEDM 2025, reporting the first integration of a silicon‑based GaN low‑voltage power‑amplifier MMIC into a handset with over 50% efficiency, marking a breakthrough for 6G‑era RF front‑ends.
The MEET 2026 conference showcased Daniel Povey’s analogy of AI evolution to biological evolution, Xiaomi’s open‑source AI breakthroughs such as Zipformer and Zapformer, and the company’s multi‑agent automotive quality engine that leverages large‑scale models, data‑driven diagnostics, and open collaboration to accelerate intelligent technology across industries.
AAAI 2026 accepted seven Xiaomi research papers—two oral presentations—covering multimodal sound editing, embodied 3D agent scheduling, scalable Text-to-SQL schema linking, parallel speculative decoding, long‑form speech QA, high‑level spatial navigation, and VLM‑driven autonomous‑driving adversaries, each with concrete datasets, methods, and benchmark gains.
The Xiaomi automotive research team secured first place in the ICCV 2025 RealADSim New View Synthesis challenge by presenting a four‑stage, 3D Gaussian Splatting‑based pipeline that tackles LiDAR‑free initialization, structured road modeling, generative pseudo‑ground‑truth refinement, and time‑invariant adaptation, achieving top PSNR, SSIM, and LPIPS scores.
