M4 pushes local AI mainstream
- Apple continues broadening on‑device AI by positioning M4 across mainstream notebooks and tablets. - Reviews note M4 configs up to 10 CPU cores, 10 GPU cores, and up to 32GB unified memory. - That tiered strategy prioritises acceptable local inference within thermals and product segmentation rather than chasing raw datacenter benchmarks ( ).
Apple has moved its M4 chip from premium demos into everyday Macs and iPads, putting on-device artificial intelligence in mainstream hardware. (apple.com) On a device, “local” artificial intelligence means the model runs on the laptop or tablet itself instead of sending every request to a remote data center. Apple says Apple Intelligence first checks whether a task can be completed on device and sends only more complex requests to its Private Cloud Compute servers when needed. (apple.com) The MacBook Air Apple introduced on March 5, 2025 starts with 16 gigabytes of unified memory and can be configured with an M4 chip that has a 10-core central processing unit, up to a 10-core graphics processing unit, and up to 32 gigabytes of unified memory. Apple says the notebook also supports Apple Intelligence in macOS Sequoia. (apple.com) Apple extended that same positioning to the iPad Air on March 4, 2026, when it updated the tablet to M4. Apple’s current specifications list an 8-core CPU, 9-core GPU, 16-core Neural Engine, 12 gigabytes of unified memory, and 120 gigabytes per second of memory bandwidth for the new iPad Air. (apple.com, apple.com) The chip itself was introduced in May 2024 with Apple’s claim of a 16-core Neural Engine capable of 38 trillion operations per second. Apple also said M4 could be built with up to 10 CPU cores and a 10-core GPU, figures that later became the ceiling for mainstream M4 Macs. (apple.com) This is a different target from cloud artificial intelligence systems built around large server clusters. Apple’s privacy documents say Apple Intelligence routes some jobs to the cloud only when a larger model is required, which means the company is designing around a split workload rather than trying to fit every generative task onto a fanless notebook or tablet. (apple.com, apple.com) The hardware choices show that balance. MacBook Air keeps the M4 family’s 120 gigabytes-per-second memory bandwidth and tops out at 32 gigabytes of unified memory, while higher-end M4 Pro systems go to 64 gigabytes and 273 gigabytes per second, giving Apple room to separate everyday local inference from heavier professional workloads. (apple.com, apple.com) Reviewers have framed the new devices less as record-setting artificial intelligence machines than as products with enough headroom for practical local features. T3’s review of the 2026 iPad Air called it “pro performance at an amateur price,” while Pune Mirror highlighted MacBook Air M4 configurations with up to 32 gigabytes of memory in India. (t3.com, punemirror.com) Apple’s bet is that mainstream buyers do not need data-center-class chips to summarize notes, rewrite email, search photos, or run lighter models privately on a personal device. By spreading M4 across the MacBook Air and iPad Air, Apple is making that level of local artificial intelligence a standard feature instead of a niche upgrade. (apple.com, apple.com)
