Draft poster · the finished cut will appear here
What happened
The Verge describes an early benchmark pass comparing an M5 Ultra Mac Studio with a similarly configured M3 Ultra. The review unit is listed as having a 36-core CPU, an 80-core GPU, 256GB of unified memory, and 4TB of storage. The Verge frames the machine for AI developers and demanding 3D and visual-effects work.
How it works
The point of the Ultra configuration is the combination of parallel compute and a large shared memory pool. The Verge reports stronger CPU results, GPU gains in several tests, fast storage, and improvements in media and 3D workloads. The author had used the machine for only a handful of days and had not completed a deep local-AI evaluation.
That makes the headline a direction of travel, not a complete buying conclusion. A result from a very large configuration may matter to a specific workload while saying little about a smaller system or a task that cannot saturate the hardware.
Why it matters
Ultra-class hardware pays off when software can use parallel compute and a very large unified-memory pool. The context around the benchmark is as important as the lead: workload, configuration, and test depth decide who benefits.
What we don’t know yet
The Verge’s report is an initial-impressions pass, not a full review. It does not establish performance across configurations, sustained thermal behavior, or local-AI performance, and a result from the largest configuration may not predict the smaller models.
Source: The Verge, Antonio G. Di Benedetto, September 21, 2026 - original review