Arm Blog

Discover how Arm Cortex-M and Cortex-R processors provide the real-time performance, reliability and safety foundations for next-generation robotics.

Help shape the future of Arm's mobile game profiling tools. Share your Streamline workflows, essential features, and challenges in our developer survey

Discover how Studio Atelico brings Bobium Brawlers to Android with Arm SME2, running generative AI on-device for creature creation, card decks, and battle

Discover how Arm Ethos-U85 and ExecuTorch enable diffusion-based face generation on Alif Ensemble E8, producing 128 x 128 images on embedded hardware

Explore when lookup-table GEMM can outperform traditional multiplication for ultra-low-bit LLMs, and how model shapes, packing, and hardware affect result

This blog series introduces the Arm FEAT_FP8 features, detailing the rationale behind FP8 data formats, their ecosystem momentum, and hardware support in Arm architecture

In Part 2, explore FP8 instructions, from conversion and scaling to dot products, matrix multiplication, and outer products in Advanced SIMD, SVE, and SME

Discover how Everysight uses Arm Ethos-U85 and on-device AI to power gaze-based interactions in lightweight smart glasses without compromising wearability

Natural-language AI helps manufacturers make advanced appliance features easier to use, creating more intuitive experiences and helping products stand out

Explore Arm security features for Linux, including PAC, BTI, and GCS, with practical steps to check CPU support, configure protection, and diagnose faults

FEAT_MOPS引入专门的memcpy, memset指令，让软件通过简单三条指令完成任何数据大小，地址对齐的memcpy, memset操作。同时让CPU微构架实现硬件自己来优化内存访问过程，让memcpy, memset这类软件中最常用的操作标准化,简单高效。

This blog post introduces Arm SME2 quarter-tile outer product instructions (FEAT_SME_MOP4) and structured sparsity outer product instructions (FEAT_SME_TMOP)

Move ROS 2 robotics from simulation to real Wi-Fi on Arm using Zenoh to keep sensor data complete and usable across devices and constrained wireless links

Discover how opacity micromaps accelerate ray-traced foliage shadows on Arm Mali G2, with Moku benchmarks showing performance gains of up to 31% in scenes

Arm Performance Libraries (ArmPL) 26.07 includes a host of useful updates.  These include new BFloat16 BLAS routines, new matrix addition and sparse matrix solving functionality, and performance tuning…

Explore the role of vector databases in agentic AI applications and how Arm Neoverse platforms deliver the performance and efficiency needed to support AI retrieval workloads at scale

Use the Arm AI Model Optimizer to optimize, evaluate, and compare your models so you can build your best AI applications on Arm

Learn how ExecuTorch brings PyTorch deployment to Arm Cortex-M CPUs with CMSIS-NN acceleration, low memory use, and support for embedded ML workflow today

Co-authored by Jason Zhu, Jackie Lee, and Damien Dooley (Arm), and Greyson Xu, Neo Zhang, and Yiming Li (Tencent Media Lab)

Explore how Arm adapters for Google Model Explorer help developers visualize AI models, inspect graph transformations, and identify performance bottlenecks

See how Arm engineers optimize streaming ASR with ExecuTorch and SME2, comparing model precision, word error rates and real-time performance on a Pixel 11

Explore how CMSIS-Ethos-U, the SDS Framework, and CMSIS Developer Assistant simplify Edge AI deployment, data replay, integration and debugging on devices

Explore how ExecuTorch, Zephyr and Arm Ethos-U55 bring PyTorch models to life on the Renesas RA8P1 through a practical camera-based edge AI application

Python 3.15 strengthens native development on Windows on Arm with broader package support, improved tooling, and native GitHub Actions workflows for teams

Celebrate 25 years of Buildroot and see how it helps Arm developers create reproducible embedded Linux systems for devices and intelligent edge workloads

Expanding Arm Learning Paths for developers and the AI agents they work with

Discover how the OCP RAS API standardizes fault management interfaces to reduce integration complexity across heterogeneous data center environments today