苹果公布全产品线端侧 AI 能力矩阵:iPhone 18 Pro 等最高跑 140 亿参数激活模型Apple unveils on-device AI capability matrix across full product line: iPhone 18 Pro and others run up to 14B-parameter activation models


AI free token/AI免费积分

IT之家 9 月 25 日消息,@aaronp613 今天(9 月 25 日)在 X 平台发布推文,报道称 Jamf 使用者大会(JNUC)中,苹果公司公布一张端侧 AI 推理能力比较图表,展示从 iPhone、iPad 到 Mac Studio 集群完整硬件梯队的本地 AI 上限。

JNUC 是 Jamf 每年举办的全球使用者大会,汇聚 Apple IT 管理员、行业领袖和工艺爱好者,分享设备管理、安全、自动化等维度的最新工艺与实践。

苹果公布全产品线端侧 AI 能力矩阵:iPhone 18 Pro 等最高跑 140 亿参数激活模型

苹果公布全产品线端侧 AI 能力矩阵:iPhone 18 Pro 等最高跑 140 亿参数激活模型

在今年 JNUC 活动中,苹果公司在其“IT 行业最新动态”环节展示了诸多合作内容,在聚焦 AI 端侧的演讲中展示了图片,IT之家梳理相关信息如下:

机型最大统一内存内存带宽可承载最大激活参数设备定位 / 适用情形
iPhone / iPad16 GB76 GB/s最高 140 亿轻量端侧 AI:Siri、文字润色、图片处理等日常任务
MacBook Air32 GB153 GB/s最高 350 亿便携轻薄本:中等 AI 负载,移动办公情形
Mac mini64 GB307 GB/s最高 700 亿桌面入门:更强本地模型,隐私敏感任务
MacBook Pro128 GB614 GB/s最高 1,200 亿专业级:大型模型、重度创作与开发
Mac Studio512 GB1.2 TB/s最高 4,800 亿工作站:高强度 AI 推理、影视 / 科研算力
Mac Studio 集群2 TB1.2 TB/s最高 1.6 万亿集群级:超大模型本地训练与推理

苹果这套思路依托统一内存架构,是达成设备端大模型推理的核心。内存容量、带宽直接决定硬件可加载模型的参数量,设备定位越高,本地 AI 能力越强。

苹果公布全产品线端侧 AI 能力矩阵:iPhone 18 Pro 等最高跑 140 亿参数激活模型

2026 苹果秋季新品发布会专题

本文来源:https://www.ithome.com/1/007/155.htm
若有版权问题,请来信告知:13951431913#139.com 或在右方提交评论。

IT之家, September 25 news. @aaronp613 posted on X platform today (September 25), reporting that at the Jamf Nation User Conference (JNUC), Apple released a chart comparing on-device AI inference capabilities, showing the local AI ceiling across the complete hardware lineup from iPhone and iPad to Mac Studio clusters.

JNUC is Jamf's annual global user conference, bringing together Apple IT administrators, industry leaders, and technology enthusiasts to share the latest technologies and practices in device management, security, automation, and more.

Apple unveils on-device AI capability matrix across full product line: iPhone 18 Pro and others run up to 14B-parameter activation models

Apple unveils on-device AI capability matrix across full product line: iPhone 18 Pro and others run up to 14B-parameter activation models

At this year's JNUC event, Apple showcased a number of collaborations in its "What's New in the IT Industry" segment, and displayed images during a talk focused on on-device AI. IT之家 has summarized the relevant information as follows:

Apple's approach relies on the unified memory architecture, which is the core of achieving on-device large model inference. Memory capacity and bandwidth directly determine the parameter count of models that can be loaded on the hardware. The higher the device's positioning, the stronger its local AI capabilities.

Apple unveils on-device AI capability matrix across full product line: iPhone 18 Pro and others run up to 14B-parameter activation models

2026 Apple fall new product launch event special topic.

暂无评论,快来发表第一条评论吧!No comments yet. Be the first!

📮 需求咨询📮 Inquiry