Arm C2 CPU cluster:
CPU orchestration for agentic AI
AI Summary
Agentic AI reinforces the foundational role of the CPU. Beyond executing application code, the CPU coordinates AI workflows that retrieve context, reason, and act across the compute system.
The Arm C2 CPU cluster is the CPU foundation of the Arm CSS for Mobile 2. It combines the new Arm C2-Ultra for high single-thread performance, C2-Pro for sustained efficiency, and two Arm Scalable Matrix Extension 2 (SME2) units for efficient local AI execution. Together, they orchestrate and run latency-sensitive mobile AI workloads across planning, prompt construction, local execution, and workload placement across CPU, GPU, AI accelerators, and cloud resources.
Built for responsive agentic AI
-
New flagship CPU performance:
15% higher single-thread performance and 12% faster application launch than the previous generation.1
-
Responsive agentic AI:
Up to 1.7x faster execution for AI models supporting voice, personal memory, and reasoning, with the C2 CPU cluster and SME2 helping orchestrate responsive on-device agents.1
-
Efficient local AI:
Supports up to two SME2 engines to deliver up to 70% performance uplift vs C1 CPU cluster.1
-
Sustained intelligence:
C2-Ultra enables foreground responsiveness, while C2-Pro enables sustained performance across longer-running AI loops.

The technologies powering AI orchestration
C2-Ultra: New flagship performance
C2-Ultra is the new flagship core in the C2 CPU cluster and Arm’s most powerful mobile CPU to date. It delivers high single-thread performance for application launch, UI responsiveness, agent orchestration, and foreground AI tasks where latency directly shapes the user experience.
C2-Pro: Built for sustained performance
C2-Pro delivers energy-efficient performance across everything from everyday applications to the latest agentic workflows, maximizing battery life while maintaining responsive system performance. The C2-Pro CPU leverages the same power-efficient microarchitecture as C1-Pro while implemented with the latest process technologies available on the market.
SME2: Built for local AI execution
Scalable Matrix Extension 2 accelerates the smaller, specialized models that agentic AI depends on, including speech, language, planning, reasoning, and lightweight inference. Running these workloads on the CPU helps reduce data movement and latency.
Build AI for Arm hardware
Discover pre-optimized models, agent-ready code and workflows, and optimization tools that help you take advantage of Arm hardware capabilities and build AI applications from cloud to edge.
Explore the CSS for Mobile 2 platform

Arm CSS for Mobile 2
The Arm AI-native compute platform that combines the C2 CPU cluster, Mali G2-Ultra NX, SI L2, software, and tools to enable responsive agentic AI and neural graphics.

Arm Mali G2-Ultra NX
The Arm AI-enabled mobile GPU that brings Arm Neural Technology into the graphics pipeline to deliver desktop-class visual experiences within mobile power budgets.
Talk with an expert
Learn how the Arm C2 CPU cluster can help you design AI-native mobile devices with responsive agentic AI, sustained performance, and efficient on-device intelligence.
Latest news and resources
Frequently asked questions
What is the Arm C2 CPU cluster?
The Arm C2 CPU cluster is the AI orchestration engine within Arm CSS for Mobile 2. It combines C2-Ultra, C2-Pro, and SME2 to coordinate responsive, on-device AI workloads.
What is the difference between C2-Ultra and C2-Pro?
C2-Ultra is optimized for peak AI responsiveness and single-thread performance, while C2-Pro balances performance and efficiency for sustained AI workloads.
How does the Arm C2 CPU cluster differ from the C1 CPU cluster?
The Arm C2 CPU cluster is designed for the AI-native era to better support agentic AI workloads. It introduces the new C2-Ultra CPU, enhanced SME2 capabilities, and improved AI orchestration to coordinate continuous, low-latency AI workloads across the compute system. Compared with the previous generation, the C2 CPU cluster delivers 15% higher single-thread performance and 12% faster application launch, helping enable more responsive on-device AI experiences. 1
Compared to previous generation. Testing done internally, March to August 2026; actual results may vary.