Apple drops M4 MacBook Pros built for local AI

Apple just killed the infamous 8GB starting RAM on its Pro laptops, giving the lineup the AI-era baseline it desperately needed.

Apple introduced the updated MacBook Pro family powered by the M4, M4 Pro, and M4 Max chips. Available in 14-inch and 16-inch sizes, the laptops start at $1,599 and are slated to hit shelves on November 8.

The headline for everyday users is that the base 14-inch model now starts with 16GB of unified memory out of the box, with support up to 32GB, three Thunderbolt 4 ports, and a price tag that didn't go up.

Serious silicon for local models

Under the hood, the entire M4 family is built on second-generation 3-nanometer tech. Apple claims it packs the world's fastest CPU core, backed by an updated Neural Engine and faster memory bandwidth specifically tuned to handle on-device LLMs and Apple Intelligence.

The M4 Pro turns up the wick with up to a 14-core CPU, a 20-core GPU, and a massive 75 percent boost in memory bandwidth over the previous generation. Apple was quick to point out that this bandwidth is double what you will find on competing AI PC chips.

For extreme workloads, the top-tier M4 Max supports up to a 16-core CPU, a 40-core GPU, over 0.5TB/s of memory bandwidth, and up to 128GB of unified memory. That ceiling lets developers load and prompt language models with nearly 200 billion parameters entirely on device without relying on the cloud.

The chassis also gains welcome physical upgrades. M4 Pro and M4 Max configurations now feature Thunderbolt 5 ports capable of hitting transfer speeds up to 120 Gb/s. All models can be configured with an optional nano-texture display that cuts down glare and pushes SDR brightness up to 1,000 nits outdoors. Battery life climbs to an absurd 24 hours, alongside an upgraded 12MP Center Stage webcam with Desk View support.

Why it matters

For builders, researchers, and creative pros, this release makes the entry-level Pro actually feel "pro" again without forcing an immediate $200 memory tax. If you run local models, compile heavy Xcode projects, or render in Blender, jumping from an Intel or M1 machine will feel night and day. The ability to load massive models locally on an M4 Max with 128GB of unified memory turns a portable laptop into a serious prototyping workstation.

For the broader industry, Apple just challenged the current crop of Windows AI PCs on their weakest spec: unified memory bandwidth. While competitors are marketing neural processing units with relatively modest memory pipelines, Apple is relying on raw, wide bandwidth pipelines that actually matter for running large local models.

In our view, this was a smart and overdue move. Raising the baseline to 16GB preserves Apple's credibility with technical users and ensures millions of new machines can actually run the incoming Apple Intelligence suite smoothly under macOS Sequoia 15.1. High-bandwidth unified memory looks like it has officially transitioned from a luxury spec to an absolute table-stakes requirement.