Glossary term

NPU

Neural processing unit: a chip, often inside phones and laptops, designed to run AI workloads efficiently.

How it works

A neural processing unit is a small, power-efficient block built into a phone or laptop chip to run AI models locally, for tasks like photo processing, transcription and on-device assistants. It trades raw power for low energy use.

Inside a modern device chip

NPU: AI workloads
GPU: graphics and parallel work
CPU: general tasks
Shared memory

Why it matters

NPUs make on-device AI practical without draining the battery or sending data to the cloud.

Related terms

On-device AI Edge AI GPU

Recent signals mentioning it

Price changeBiggest1 h ago · Qwen

Qwen3.8 27B input price cut 73%: $0.15 → $0.04 per 1M tokens

Input price$0.04 per 1M tokens-73%Output price$2.3 per 1M tokens+23%

Price change1 h ago · Z.ai

GLM 5.3 Flash input price cut 17%: $0.024 → $0.02 per 1M tokens

Input price$0.02 per 1M tokens-17%

Price changeBiggest6 h ago · MiniMax

MiniMax M3.1 Flash input price raised 100%: $0.075 → $0.15 per 1M tokens

Input price$0.15 per 1M tokens+100%Output price$0.6 per 1M tokens+100%

Price changeBiggest7 h ago · Tencent

Hy3 input price raised 60%: $0.083 → $0.132 per 1M tokens

Input price$0.132 per 1M tokens+60%Output price$0.528 per 1M tokens+60%