AH

Arm Kleidi

Developer tool

by Arm Holdings

Arm's AI software acceleration libraries and framework integrations (PyTorch, ExecuTorch, llama.cpp and others) that speed up model inference on Arm CPUs without extra developer work; Arm Compute Library is one of its components.

Find alternatives to Arm Kleidi

What it does

• Model inference

Sources

Pricing

Official documentation · 18 Oct 2026

The page has a section headed Open Source Technology; no price or plan is mentioned anywhere on the page.

Description

Official documentation · 18 Oct 2026

"Through Kleidi libraries, framework integrations, and ecosystem partnerships, Arm delivers seamless acceleration across Arm hardware."; "Arm Kleidi technologies are integrated into popular frameworks, accelerating models on Arm CPUs without extra developer effort."

What it does

Official documentation · 18 Oct 2026

"Integrating Kleidi improves Llama model inference by up to 18 times, Gemma 2 2B by 15 times, and performance for natural language processing (NLP) models, including 2.2 times uplift on Bert-Large."

Sold within

Official documentation · 18 Oct 2026

Arm Kleidi is a named Arm software line with its own page and no parent product; obtained as open-source libraries through frameworks and "Developer Resources".

Name

Official documentation · 18 Oct 2026

Page title "Kleidi – Software-Level AI Acceleration – Arm®" and heading "Arm Kleidi"; the Arm Compute Library page calls itself a key component of Arm Kleidi.

URL

Official documentation · 18 Oct 2026

The live product page that this record links to (Arm Kleidi).

Deployment

Official documentation · 18 Oct 2026

"accelerating models on Arm CPUs": the libraries run inside the customer's own frameworks on the device or server.

Interfaces

Official documentation · 18 Oct 2026

"Through Kleidi libraries, framework integrations, and ecosystem partnerships".

Intended for

Official documentation · 18 Oct 2026

"Seamless AI Acceleration for Developers Everywhere"; "Developer Enablement".

Also from Arm Holdings

• Arm AGI CPU Hardware

  Arm's first self-designed, TSMC-manufactured production CPU — up to 136 Neoverse V3 cores on 3nm — built for sustained agentic-AI data-centre workloads and co-developed with lead customer Meta.

• Arm Ethos-U85 Hardware

  Neural processing unit designs licensed to chip makers for running models on edge and mobile devices.

• Arm Ethos-U55 Hardware

  A licensable microNPU IP core that adds on-device ML inference to area- and power-constrained embedded and IoT designs paired with Cortex-M processors.

• Arm Ethos-U65 Hardware

  A licensable microNPU IP core delivering roughly double the on-device inference performance of Ethos-U55, for Cortex-A, Cortex-R and Neoverse-based edge systems.

• Arm Lumex CSS Hardware

  Arm's AI-first mobile Compute Subsystem, a licensable CPU, GPU and system-IP bundle for on-device generative AI, shipping in the vivo X300 and OPPO Find X9.

• Arm Neoverse Hardware

  CPU designs licensed to chip makers for servers and data-centre silicon, including the host processors in AI systems.

• Arm Zena CSS Hardware

  A licensable physical-AI Compute Subsystem combining Cortex-A and Cortex-R processors with system IP for safety-relevant AI-defined vehicle and robotics SoCs.

• Arm CSS for Mobile 2 Hardware

  Arm's second-generation mobile compute subsystem — the C2 CPU cluster (C2-Ultra, C2-Pro) with SME2 for on-device agentic AI — positioned by Arm against "the previous generation Arm (Lumex) reference platform".

• Arm Neoverse CSS N4 Hardware

  A configurable, pre-validated compute subsystem built on Neoverse N4 cores (8–128 cores per die, Armv9.3, up to 3.8 GHz on 3 nm) with integrated system IP, delivered to partners as RTL for agentic-infrastructure silicon.

• Arm C1-Ultra Hardware

  Arm C1-Ultra is a flagship CPU core that supports AI inference.

• Arm C1-Premium Hardware

  Arm C1-Premium is a CPU for mobile SoCs and edge AI inference with SME2 acceleration.

• Arm C1-Pro Hardware

  Arm C1-Pro is an Armv9.3-A CPU core that supports on-device AI inference.

• Arm C1-Nano Hardware

  Arm C1-Nano is a CPU core for always-on AI and lightweight inference.

• Arm C2 CPU cluster Hardware

  Arm C2 CPU cluster provides CPU orchestration and local execution for agentic AI.

• Arm Mali G2-Ultra NX Hardware

  Arm Mali G2-Ultra NX is a mobile GPU with integrated AI acceleration for neural graphics.

• Arm Cortex-M52 Hardware

  A processor IP design that uses Arm Helium technology for digital signal processing and machine learning in embedded devices.

• Arm Cortex-M55 Hardware

  A processor IP design for endpoint artificial intelligence, machine learning, and digital signal processing.

• Arm Cortex-M85 Hardware

  A processor IP design that uses Arm Helium technology for endpoint machine learning and digital signal processing.

• Arm Compute Library Developer tool

  Provides machine learning functions optimized for Cortex-A and Neoverse CPUs.

• Arm Corstone-300 Hardware

  Reference package for building IoT and embedded SoCs around the Cortex-M55 processor, Arm's most AI-capable Cortex-M, with the Ethos-U55 microNPU as an integrable option for on-device machine learning.

• Arm Corstone-310 Hardware

  Reference package and example subsystem pairing the Cortex-M85 CPU with the Ethos-U55 NPU, for secure, AI-capable IoT SoCs with high-performance DSP and ML.

• Arm Corstone-315 Hardware

  Reference package and example subsystem integrating the Cortex-M85 CPU, DMA-350, Mali-C55 ISP and the second-generation Ethos-U65 NPU, for AI-capable IoT devices with camera input.

• Arm Corstone-320 Hardware

  Reference package and example subsystem integrating the Cortex-M85 CPU, DMA-350, Mali-C55 ISP and the third-generation Ethos-U85 NPU with transformer-network support, for AI SoC development.

Open Arm Holdings in the directory

Something wrong here? Send a correction — quote this product id: arm-kleidi.