Apple’s A20 Pro chip has been shown to run a 27B parameter AI model at double the speed of the A19 Pro chip found in the iPhone 17 Pro. This performance boost comes from the A20 Pro’s enhanced neural processing unit (NPU), which can handle 65 TOPS (tera operations per second) compared to the A19 Pro’s 32 TOPS.
The A20 Pro’s NPU achieves this by leveraging a more efficient architecture and higher clock speeds. In practical terms, this means the A20 Pro can process complex AI tasks, such as running large language models, much faster. For instance, it can handle a 27B parameter model at 50 tokens per second, whereas the A19 Pro manages only 25 tokens per second.
This matters for mini PC and local AI users because it shows that Apple’s latest chip can handle intensive AI workloads more efficiently, making it a strong contender for devices that need to run AI models locally without relying on cloud services.
