OtherAvailable now on iPhone 17, Apple M series devices, and flagship Android phones
SME2
Specifications
- Fps
- 30
- Lux
- 1
- Power (w)
- 1
- Speed up factor
- 5
- Pipeline speed up factor
- 2.5
- Inference speed up factor
- 6
- Vision audio speed up factor
- 3
Overview
SME2 is the latest CPU extension on Arm Lumex CSS, designed to accelerate matrix-oriented compute workloads directly on device. It improves performance for AI and ML models, especially those relying on matrix multiplication, common in transformers, CNNs, and LLMs. It delivers up to 5x AI performance uplift, fully on-device inference, and seamless framework support via Arm KleidiAI, with cross-platform availability on iPhone 17, Apple M series, and flagship Android phones.
Key features
- ▸5x AI performance uplift vs previous generation
- ▸Fully on-device AI inference for privacy and low latency
- ▸Seamless framework support via Arm KleidiAI (PyTorch, LiteRT, etc.)
- ▸Cross-platform ready on iPhone 17, Apple M series, flagship Android
- ▸Low precision neural network support (4-bit and 2-bit)
- ▸Scalable from entry-tier to flagship mobile devices
- ▸Up to 6x faster LLM inference, 3x improvement in vision/audio
- ▸Integrates with PyTorch, ONNX Runtime, XNNPACK without code changes