Guozhen AIGlobal AI field notes and model intelligence

Realtime AI News

MediaTek launches Dimensity 9600 Pro with 2nm process and AI-native architecture for on-device agents

MediaTek has released the Dimensity 9600 Pro, a flagship mobile platform built on a 2nm process and using an AI-native architecture aimed at running AI agents on the device itself. The launch, reported by Electronic Engineering Album, puts on-device intelligence rather than cloud inference at the centre of the chip's pitch.

Published
联发科发布天玑9600 Pro:2nm制程与AI原生架构推动端侧智能体落地
Image source: mediatek.com

MediaTek has released the Dimensity 9600 Pro, a new flagship mobile platform that puts on-device AI agents at the centre of its product story. According to a report by Electronic Engineering Album, the chip is built on a 2nm process and uses an AI-native architecture.

The launch rests on two technical claims. The first is the process node: 2nm is the most advanced generation of mass-production silicon the industry currently talks about, and it is generally expected to fit more transistors into a similar area, giving designers more room to trade performance against power. The second is the architecture: an AI-native design suggests that compute and memory paths were reserved for always-on AI workloads from the start, rather than bolted on later through software.

The on-device agent is the explicit target. Running models locally means inference and task execution happen on the phone rather than being shipped to a cloud server for every request. The standard arguments for that approach are lower latency, offline availability and the ability to keep personal data on the device.

That has become one of the main battlegrounds in mobile silicon. As model compression, quantization and smaller inference stacks mature, capabilities that once lived only in data centres are moving to handsets, which is why chipmakers, phone makers and model developers now work more closely than they did in earlier generations.

By framing the Dimensity 9600 Pro around agents, MediaTek is betting on software that perceives, plans and acts in the background rather than answering a single question. The bet is that agents should run where the user is, not in a distant data centre. Those workloads stress different parts of a chip: sustained efficiency, memory bandwidth and thermal stability matter as much as peak compute.

The release information available so far stays at the level of process and architecture. Benchmark numbers, the first devices to carry the chip and its commercial timeline have not been detailed in the available report, so any assessment of the platform will have to wait for independent testing.

Whether on-device agents become a mainstream experience depends on chips being able to keep inference and task execution on the handset without wrecking battery life. The Dimensity 9600 Pro is the latest sample of that shift, and how many agent applications it actually attracts is the thing to watch.

Why it matters

MediaTek is tying 2nm silicon to an AI-native design in the same generation, a sign that mobile chip competition is shifting from raw compute toward the efficiency and local inference that on-device agents require. If the experience lands, it will reshape both handset AI features and the criteria buyers use to judge a flagship chip.

MediaTekChipOn-device AI
Back to realtime news

Nearby Updates

All

09/17, 07:56

Google DeepMind Sees Early Signs of Recursive Self-Improvement Ahead of Gemini 4

Google DeepMind staffer Logan Kilpatrick said the company is seeing early signs of recursive self-improvement, with stronger models speeding up the development of the models that follow them, and he added that other frontier labs are seeing the same thing. He pointed to Gemini's three-to-four-week release cadence and said the unreleased Gemini 4 will be Google's largest and most ambitious pre-training run so far.

09/17, 08:41

AI may need aviation-style safety rules to earn trust, Josephine Teo says

Singapore minister Josephine Teo said AI may need aviation-style safety regulation before the public can trust it. The comparison reported by The Straits Times centres the debate on verifiable safety mechanisms and clear accountability rather than capability races.

09/17, 08:55

Naver AI Shopping Agent Sees 82% Transaction Surge, Adds Delivery-Based Recommendations

Naver's AI shopping agent recorded an 82% transaction surge and added delivery-based recommendations, according to Chosun Ilbo. The disclosure suggests conversational shopping assistants are starting to carry real orders in Korea, not just improve the browsing experience.

09/17, 06:34

A safety push sparks infighting at OpenAI and Anthropic, AFR reports

A push on AI safety is fuelling internal friction at OpenAI and Anthropic, according to The Australian Financial Review. The framing puts both frontier labs in one story: public safety commitments keep expanding while disagreements over direction and pace surface inside.