Arm CSS for Mobile 2 delivers advanced AI-powered mobile experiences across the ecosystem  

From more responsive on-device AI to cinematic gaming, the next generation of AI-powered mobile experiences is taking shape. Launched in September 2026, Arm CSS for Mobile 2 is the compute foundation for this new era of mobile AI. The AI-native compute platform combines the Arm C2 CPU cluster with Arm Scalable Matrix Extension 2 (SME2) and Arm Mali G2-Ultra NX, Arm’s first Mali GPU featuring Arm Neural Technology, along with Arm system technologies and software.  

The Arm C2 Ultra CPU with SME2 brings a new level of responsive, efficient on-device AI to the CPU, while Mali G2-Ultra NX integrates dedicated neural accelerators that enable neural graphics to run natively on the GPU. These CPU and GPU technologies are already being adopted across the mobile ecosystem by leading OEMs, and game engines and studios, highlighting the platform’s ongoing momentum. 

Bringing Arm’s latest CPU and GPU technologies to new mobile devices 

The capabilities of the C2 CPU cluster and Mali G2-Ultra NX are coming to life on flagship smartphones built on MediaTek’s Dimensity 9600 Pro. vivo’s X500 Pro and Pro Max smartphones – the first Dimensity 9600 Pro-based flagship devices on the market – are designed to deliver multitasking and sophisticated agentic AI experiences. The OPPO Find X10 Pro Max brings faster performance and richer visuals to the Find X10 Series, which focuses on AI, photography, and gaming.  

Separately, Xiaomi’s XRING O3 chipset brings Mali G2-Ultra NX graphics to the Xiaomi 18 Fold smartphone and Xiaomi Pad 9 Pro Max tablet. 

Mali G2-Ultra NX brings AI-native graphics to mobile gaming

Mali G2-Ultra NX is built for neural graphics from the ground up, with neural accelerators tightly integrated into the shader codes. This allows neural graphics workloads to run alongside graphics and compute workloads.  

Mali G2-Ultra NX combines three neural technologies to deliver sharper graphics, smoother gameplay, and cinematic lighting within mobile power budgets: 

• Neural Super Sampling (NSS), which reconstructs higher-resolution images from lower-resolution renders, delivering sharper visuals with less rendering work. 

• Neural Frame Rate Upscaling (NFRU), which generates intermediate frames to increase frame rates and deliver smoother gameplay. 

• Neural Super Sampling and Denoising (NSSD), which combines neural upscaling with denoising to deliver cleaner, more detailed visuals in complex ray-traced scenes. 

Players can experience these capabilities firsthand in Neural Dawn, co-developed by Arm and Sumo Digital. The game is now available to download and play on supported Android devices powered by Mali G2-Ultra NX, including those highlighted earlier. It is also the first mobile game featuring Unreal Engine MegaLights, bringing hundreds of dynamic lights and ray-traced shadows to mobile.  

Meanwhile, leading game engines and studios are integrating Arm Neural Technology into their graphics pipelines, extending the progress beyond Neural Dawn.  

• Unity China’s Tuanjie Engine version 1.11.0 integrates NSS, giving developers access to neural graphics through an established engine and reducing the need for custom implementations. To date, the Tuanjie Engine has already surpassed 1.9 million cumulative downloads, laying a strong foundation for the broader adoption of Arm Neural Technology. 

• Infold Games is integrating NSS into Infinity Nikki, which has surpassed 20 million downloads worldwide. 

• The MagicDawn team at Tencent Games Central Tech is integrating Arm Neural Technology into its MagicDawn rendering technology.

• A NSSD demonstration in Arena Breakout Infinite, which has attracted more than 66,000 concurrent players on Steam, showcases how AI-assisted graphics can support sophisticated game content. 

• NetEase is bringing Arm Neural Technology to players through its Where Winds Meet game, which is a Gamescom Best Mobile Game award winner. 

SME2 is accelerating AI experiences on mobile

Through SME2 in CSS for Mobile 2, the performance of agentic AI workloads is improving by up to 70% compared to the previous generation. Leading smartphone vendors like OPPO and vivo are using SME2 to accelerate on-device AI workloads. Google is helping developers use SME2 through its AI Edge software stack, where LiteRT, XNNPACK, and Arm KleidiAI enable automatic CPU acceleration on supported mobile devices. 

Internal Arm testing shows how SME2 delivers significant performance gains when running AI workloads on Google’s Gemma 4 E2B model, including an average of 5.5x speedup in prefill (processing user input) and up to 1.6x faster decoding (generating responses). Users can explore SME2-accelerated on-device AI through Google AI Edge Gallery, an application that runs Gemma models locally, and leverages SME2 acceleration on supported devices. 

Other collaborations on SME2 include: 

• Arm and Tencent Media Lab working together to use SME2 to accelerate neural image decoding, cutting processing time for a 2K image from 61ms to 30ms on a reference smartphone using just two CPU threads. This enables faster, more responsive image processing on mobile devices. 
• Liquid AI using SME2 to accelerate the performance of its new LFM3 models, with early, provisional tests showing 3.1x faster prefill and 1.5x faster decoding for LFM3-1.7B using two CPU threads. This helps AI-based applications process inputs and generate responses faster, improving responsiveness for on-device AI. 
• Studio Atelico using SME2 to bring the Bobium Brawlers game to Android, with this enabling faster, more responsive experiences for players through AI that runs entirely on mobile devices.  

Building momentum for the next generation of mobile experiences

The momentum behind CSS for Mobile 2 is taking shape across the mobile ecosystem. OPPO and vivo are bringing Arm CPU and GPU technologies into commercial mobile devices. Meanwhile, Xiaomi is extending the reach of Mali G2-Ultra NX across smartphones and tablets. 

These hardware innovations are translating into enhanced real-world experiences, with the software ecosystem taking advantage of the advanced compute capabilities. Game engine and studio integrations, and playable demonstrations like Neural Dawn give developers more ways to harness neural graphics, and deliver richer, more cinematic mobile gaming experiences. And the growing adoption and integration of SME2 enables faster, more responsive AI experiences across mobile applications. 

As these mobile technologies reach more devices and development environments, Arm is laying the foundation for more sophisticated on-device AI, advanced graphics, and increasingly immersive and responsive mobile experiences.