Arm Lumex Platform comes to life with AI-powered smartphones, apps and experiences

Summary:

• Arm Lumex Platform ushers in a new era of intelligent performance, combining the Arm C1 CPU cluster integrating Arm Scalable Matrix Extension 2 (SME2) and Arm Mali G1-Ultra GPU to deliver faster, smarter, and more personalized on-device AI.
• Leading smartphone makers like vivo and OPPO are already leveraging Lumex, achieving tangible AI performance gains in translation, image recognition, transcription, and speech processing.
• Developers and leading applications, including Alipay, Alibaba Taobao, Google’s apps and leading gaming apps, are transforming user experiences through utilizing SME2 and Mali G1-Ultra performance benefits.

In September 2025, Arm launched the Lumex Platform, our most advanced smartphone compute platform to date, which is designed to make artificial intelligence (AI) faster, smarter, and more personal. With the new Arm C1 CPU cluster integrating Arm Scalable Matrix Extension 2 (SME2) and Arm Mali G1-Ultra GPU, Lumex puts advanced intelligent computing into the hands of billions worldwide – across leading smartphone devices and apps that are bringing the benefits of on-device AI to life.

Powering new flagship smartphones

Just one month after its debut, the effects of Lumex are rippling throughout the tech landscape and beyond. Device makers like vivo and OPPO are among the first to harness the power of Lumex technologies, launching flagship smartphones that integrate the SME2-enabled C1 CPUs and Mali G1-Ultra GPU to provide tangible real-world improvements in on-device AI.

The vivo X300 series delivers 20 percent faster live translation and 30 percent quicker image recognition, with SME2’s low-precision matrix acceleration enabling instant, efficient and private language translation and photo sorting.

OPPO’s Find X9 series demonstrates measurable gains in automatic speech recognition through a significant performance uplift and reduction in quantization compared with previous models. This leads to more responsive voice control and transcription across everyday apps running on the flagship smartphone devices.

Even before these Android innovations, SME2 had already been making waves in iOS, accelerating on-device intelligence across a range of devices, including the iPhone 16 and 17, iPad Pro, MacMini, MacBook Pro and MacBook Air. SME2 is highly portable, with cross-platform reach through iOS and Android to benefit billions of Arm-based devices.

Transforming apps and experiences

Lumex isn’t just about faster devices, it’s about what people can do with them. Developers are utilizing SME2 to accelerate large language model (LLM) inference, image processing, and even audio generation for faster real-time experiences across a range of leading applications. As SME2 works transparently through extensive Arm KleidiAI integrations in leading frameworks and runtimes, like LiteRT, MediaPipe, MNN, ONNX Runtime, and XNNPACK, developers can build their apps and workloads once and then deploy across a wide range of Arm-based devices – no rewrites, no additional complexity.

Automatic travel video creation

For Alipay, a payment and lifestyle app, SME2 powers the automatic creation of travel videos, editing video clips and photos as users browse their gallery. Working with Alipay and vivo, Arm helped to achieve a 40 percent reduction in the time it takes for LLM responses to interact with users during this process for faster on-device generative AI. The same SME2-enabled acceleration also reduces power consumption, helping Alipay’s engineers ship complex vision and language models that run efficiently within mobile thermal and battery limits.

“SME2’s matrix acceleration and efficient low-precision computing allow us to deliver real-time AI with low latency, enhanced privacy, and seamless integration into everyday scenarios.” Xindan Weng, Head of Client Engineering, Alipay.

Faster productivity experiences

Google is already optimizing productivity and generative-AI features across its range of apps, like Gmail, Google Photos and YouTube, when they run on SME2-enabled hardware. Google’s Gemma 3 model delivers 6x faster AI responses when running on chat interactions compared to the same device without SME2 active. Also, on a single CPU core with SME2 acceleration, Gemma 3 can start text summarization of up to 800 words in under one second.

“We’re really excited about Arm’s new offering (SME2) because it brings additional compute to the CPU. For the first time, we can run these (GenAI) capabilities across a wide range of devices in the ecosystem.” Oli Gaymond, Head of AI/ML Product, Android, Google.

Personalized product recommendations

The Alibaba Taobao e-commerce app leveraged SME2 across enabled smartphones via the MNN framework to power on-device product ranking based on user preferences. Billion-parameter models like the Alibaba Qwen model now re-order results as users scroll, while keeping data private. For developers, SME2’s efficient matrix handling cuts memory bandwidth, letting them deploy large models within the same silicon envelope.

“Through deep integration with SME2, MNN enables low-latency, quantized inference for billion-parameter models like Qwen on smartphones – showcasing Arm and Alibaba’s joint innovation in scalable, next-gen mobile AI.” Xiaotang Jiang, Head of MNN, Taobao and Tmall Group, Alibaba.

More intuitive, intelligent gaming

For leading mobile gaming apps, including Fortnite and Genshin Impact, the Mali G1-Ultra GPU provides across-the-board real-world performance improvements. These are made possible by the next-generation Arm Ray Tracing Unit (RTUv2) that delivers a 2x uplift in ray tracing performance for desktop-quality lighting, reflections and shadows.

Looking further ahead, Arm neural technology will add dedicated neural accelerators to 2026 Arm GPUs. This will bring PC-quality, AI powered graphics to mobile for the first time, while laying the foundation for future on-device AI innovation. In preparation for future smartphones with Arm neural technology built-in, developers can begin deploying their own neural models and workloads fine-tuned to their gaming content now through the world’s first Neural Graphics Development Kit.

Alongside Arm GPUs, SME2 also has an impact on gaming apps, with the technology delivering 2.5x performance gains in Tencent’s GiiNEX game engine, enabling developers to build richer worlds alongside smarter real-time interactions and non-player characters (NPCs). The future potential of real-time gaming interactions is explored in Arm’s new Space Bartender demo, which utilizes KleidiAI optimizations to unlock smarter, faster, more intuitive gaming interactions with NPCs.

Powering a new era of enhanced creativity and wellness

SME2 has the potential to power a new generation of advanced creativity and wellness user experiences. The technology can be used by device manufacturers and developers to power real-time computational photography to bring sharper, more vivid images and photos to life. Meanwhile, an Arm “AI Yoga Tutor” demo for mobile devices and TVs uses AI techniques, including pose estimation, LLMs and text-to-speech, to give users real-time conversational feedback tailored to their yoga poses, with the SME2-enablement leading to 2.5x speed-ups in the AI pipeline for faster interactions. These demonstrate the potential of SME2 to deliver tangible performance improvements to a new generation of AI experiences.

Bringing on-device AI to life

Lumex is already enabling real-world impact, from the devices in people’s hands to the apps they use every day. The results are measurable: developers are delivering next-gen AI experiences faster with fewer constraints, OEMs are designing faster, more power-efficient smartphones, and users are experiencing advanced, instant AI that accelerates their lives. Lumex is defining a new standard for intelligent performance — one that brings on-device AI to life for billions of users worldwide.