Announcing Arm Performix: Empowering developers with scalable performance for the age of AI agents

Arm Performix enables developers to scale performant, efficient AI agent experiences by unifying performance insights and optimization across the Arm compute platform.

• Arm Performix is a free performance analysis toolkit for modern agentic development workflows — the first of its kind, defining a new class of performance tooling for developers and AI agents. 
• Through clear, deep performance insights, any developer or AI agent can use Arm Performix to understand, analyze and optimize applications running on Arm-based cloud platforms. 
• Microsoft, MongoDB, Redis and SAP have shared their support for Arm Performix. 

By Alex Spinelli, SVP, AI and Developer Platforms, Arm

April 28, 2026 -- Arm is delivering a new class of compute designed for the scale, efficiency and intensity of AI workloads. The historic launch of the Arm AGI CPU marked an inflection point, not only in silicon, but in how software is built, deployed and optimized for AI. With 1.25+ billion Arm Neoverse cores shipped, Arm-based compute has already reached significant scale. This growth spans from hyperscaler-designed silicon based on Arm Neoverse to next-generation platforms like Arm AGI CPU. 

Built on this foundation and further accelerating momentum on Arm-based platforms, Arm is introducing Performix, a performance analysis toolkit to help developers optimize AI systems end to end. Integrated in modern workflows, it delivers continuous, expert performance analysis and insights so developers can act faster. Arm Performix expands Arm’s history of strong performance beyond silicon and into the full software stack. 

The shift to agentic AI is introducing new performance challenges. Workloads are becoming more complex, spanning multiple components and layers, while traditional tooling remains designed for monolithic, manually analyzed applications. At the same time, performance tools must integrate into automated workflows and be usable directly by AI agents, with outputs that are structured, repeatable and actionable.

Built for modern agentic AI development

Arm Performix defines a new class of performance tooling that is a step-change from manual, fragmented traditional offerings. It is a free performance analysis toolkit for software and modern AI development. Performix gives developers and AI agents system-wide visibility to identify inefficiencies and optimize workloads running on Arm-based infrastructure, while fitting directly into automated, agent-driven workflows. 

By collecting performance data directly from Arm-based silicon at runtime, Performix turns low-level hardware behavior into guided, actionable insights in real-world environments. This enables developers to move beyond manual analysis, allowing them to confidently validate, tune and scale workloads on Arm-based infrastructure — from cloud deployments to next-generation silicon platforms like Arm AGI CPU.

Turning hardware data into actionable, accelerated insights 

Performix is an extensible platform that provides system-wide analysis across key metrics such as memory bandwidth, latency, cache efficiency and CPU utilization. What once required deep architectural knowledge is now a clear, accessible process for developers and AI agents alike. The guided, recipe-based approach helps developers shift from reactive tuning to focus on the areas that matter most, turning low-level hardware signals into clear optimization paths. 

Central to this is the Arm MCP Server, which allows developers and AI assistants to run Performix directly from tools such as GitHub Copilot, Kiro, Gemini and Codex. Analysis can be triggered from within the development environment, with results surfaced alongside the code. The Arm MCP Server contains expert knowledge and tools to enhance agentic development workflows targeting Arm-based systems, from migration to optimization. This creates a tight feedback loop where performance evaluation becomes a continuous, automated part of development. 

In 2025, 50% of the CPU compute shipped to top hyperscalers was Arm-based.¹ Arm and cloud partners have used these capabilities to identify performance bottlenecks and unlock peak performance in real workloads. This builds on Arm’s broader ecosystem momentum, where optimizing for Arm is increasingly a competitive advantage. 

Performix has been developed with feedback from partners and clients like Microsoft, MongoDB, Redis and SAP.

Continuous performance insight for AI agents

Performance depends not only on hardware; it depends on how effectively developers can optimize and scale software across complex systems. Performix unlocks the full potential of the Arm platform through highly accessible, deep performance insights, so developers spend less time interpreting data and more time optimizing where it matters. Together, Arm-based platforms—from hyperscaler deployments to Arm AGI CPU—and Performix strengthen Arm’s foundation for efficient, high-performance AI systems. This marks yet another significant step in Arm’s long-standing commitment to our ecosystem, with Performix empowering developers to build, optimize and scale the next generation of cloud and AI infrastructure on Arm. 

Learn how to use Performix in this Arm Community blog. 

Learn more and get started with Arm Performix today. 

¹ Based on internal Arm estimates.

Partner quotes:

“As developers and cloud architects increasingly seek to move existing x86 cloud workloads onto Arm, they need clear metrics and insights to help them identify performance bottlenecks efficiently. Arm Performix was developed with extensive feedback from our team to help simplify this migration,” said Pat Stemen, Vice President, Azure Hardware Systems and Infrastructure, Microsoft. “We look forward to seeing customers utilize Arm Performix to optimize their applications for Cobalt 200 virtual machines.” 

“As a senior member of the performance engineering team at MongoDB, I was extremely impressed with the depth of analysis Arm Performix provides. The tool is approachable, and quick to set up and run,” said Jawwad Asghar, Senior Software Engineer at MongoDB. “I was able to generate a wealth of data about where to target optimization efforts across our code base. The CPU Microarchitecture analysis showed a clear front-end bottleneck and which functions were the most inefficient at different cache levels. Using the actionable insights Performix generated, we found a clear path to achieving a 10% performance boost on one of our key workloads in just a few days. The tool has been a big win for developer velocity in our performance tuning process.” 

“Agentic AI workloads are driving unprecedented scale and complexity, making it critical to understand performance across the full stack,” said Filipe Oliveira, Performance Analysis and Optimization Lead, Redis. “We’re working with Arm to optimize Redis on Arm-based cloud infrastructure. Arm Performix brings system-wide visibility and actionable insights we need to identify inefficiencies quickly. Performix is an enabler to streamline and continuously optimize performance as we scale AI and data-intensive workloads.” 

“Arm Performix streamlines performance analysis using intuitive visualizations and pre-configured recipes,” said Lars Hoemke, Head of HANA Core Performance, SAP. “With SAP HANA, Arm Performix allows us to identify code hotspots, drill down into issues, compare systems quickly and gain clear performance insights that save time.”