Google Releases LiteRT Tensor API for Cross-Platform Edge Model Authoring

1 reports archived1 independent sourcesupdated 11/13/2026, 09:31:00

Synthesis & Latest Updates

1 Sources Cross-Validated

Google has launched the LiteRT Tensor API, a lightweight C++ interface that lets developers author edge model graphs operator by operator, running the same code across CPU, GPU and NPU without per-hardware rewrites. It supports direct weight import from Hugging Face safetensors, zero-copy buffer interop, and AI coding agents that generate and optimize graphs; in Google's demo, identical SAM 2 code ran about 1.5x faster than Apple MLX on a MacBook Pro.

LATEST/Google has launched the LiteRT Tensor API, a lightweight C++ interface that lets developers author edge model graphs operator by operator, running the same code across CPU, GPU and NPU without per-hardware rewrites. It supports direct weight import from Hugging Face safetensors, zero-copy buffer interop, and AI coding agents that generate and optimize graphs; in Google's demo, identical SAM 2 code ran about 1.5x faster than Apple MLX on a MacBook Pro.

Coverage timeline

Total 1 reports · Latest first

InfoQ 中文T2Nov 13, 09:31·72 pts
Google Releases LiteRT Tensor API for Cross-Platform Edge Model Authoring
- One C++ graph runs on CPU, GPU and NPU, eliminating duplicate Metal, OpenCL and Vulkan implementations
- Direct safetensors weight loading and zero-copy buffers simplify the full preprocess-to-postprocess pipeline
- Coding agents can generate, modify and benchmark graphs; SAM 2 ran about 1.5x faster than Apple MLX on a MacBook Pro
View Related Reports & Analysis Original Source