Apple today revealed the A20 Pro chip in the iPhone 18 Pro and ‌iPhone 18 Pro‌ Max, built on a brand new architecture using 2nm process technology.


Apple described the A20 Pro as "an exceptional chip that sets a new bar for mobile computing." The chip has a 6-core CPU design with two new super cores that Apple says are up to 20% faster, calling them "desktop class" and "the fastest in any smartphone," alongside four new efficiency cores.

The GPU has also been redesigned with a 7-core layout that Apple says delivers up to 40% faster graphics performance and increased memory bandwidth. New neural accelerators built into the GPU offer twice the FP8 performance of the previous generation, which Apple said helps the chip run large language models locally on the device.

The A20 Pro features an additional Neural Engine, bringing the total core count to 32. The company says the redesigned Neural Engine delivers 2x the compute of its predecessor, is more efficient, and now supports 8-bit floating point math. Apple called the A20 Pro "the ultimate chip for running advanced on-device models," adding that it ships with "the widest memory interface we've ever shipped in an iPhone."

The company said effective thermal management in the ‌iPhone 18 Pro‌ is critical to unlocking the chip's maximum performance, since efficient heat dissipation determines how much of that performance can be sustained. To support this, Apple redesigned the chip's silicon packaging, describing it as "custom packaging inspired by our M-series chips" that connects the silicon and memory directly to the vapor chamber for improved thermal performance.

Apple compared sustained performance directly against previous generations, citing 40% more sustained performance in ‌iPhone 18 Pro‌'s A20 Pro compared with the A19 Pro in the iPhone 17 Pro, and twice the sustained performance compared with the A18 Pro in the iPhone 16 Pro.
This article, "Apple Unveils A20 Pro as First 2nm Smartphone Chip" first appeared on MacRumors.com

Discuss this article in our forums

Apple told Mac App Store developers they can now ship Apple silicon-only apps, dropping support for Intel Macs.

(via Cult of Mac - Your source for the latest Apple news, rumors, analysis, reviews, how-tos and deals.)

Perplexity has launched Hybrid Compute for its Computer agent on Apple Silicon Macs, a system that splits a single task between frontier models in the cloud and compact models running locally so confidential files and personal data never have to leave the machine.

The idea is simple and, if it holds up in daily use, important: Computer still starts work in the cloud, where large models handle research, planning, and long-horizon reasoning. When a local classifier spots sensitive information — names, emails, account numbers, privileged files — the user is prompted to keep that slice of the job on the Mac, mask the details, or send it to the cloud anyway. Cloud and local pieces stay in the same workflow. The user does not have to restart the task or shuttle files between tools.

At launch, local options include Gemma 4 E4B, Qwen3.6 35B-A3B, and a Perplexity post-trained Qwen variant. Users pick the cloud orchestrator; Perplexity selects helper frontier models. Hybrid Compute is available today to Pro, Max, and Enterprise subscribers in the Perplexity Mac app on Apple silicon running macOS 15 or later, with 24GB of unified memory as the minimum and 32GB recommended. Setup is a one-click model download rather than a custom local-runtime stack.

The Mac does not have to be the device that starts the job. A prompt from iPhone or iPad can still kick off a session; if the Mac is on and running Perplexity, sensitive steps can execute there against local files and apps. That makes a always-on Mac mini a plausible private compute node for people who work from a phone most of the day.

Perplexity has been building toward this split for months. In June it described hybrid agentic inference as a way to optimize for accuracy, privacy, and token cost at once: keep financial records, health information, and personal files local; send only the work that truly needs a frontier model to the server. Portable Computer, a more local-first variant, already existed for users who want most of the stack on device and only escalate with permission. Hybrid Compute approaches the same problem from the other direction — cloud-first orchestration with an on-device privacy gate.

That gate is the linchpin. An on-device PII classifier, which Perplexity says it has open-sourced and trained with its Secure Intelligence Institute, inspects content before anything is transmitted. Credentials, payment numbers, and government IDs get the strictest treatment. For enterprises, admins can set organization-wide rules for what must stay local, what may be masked, and what requires approval, plus audit when data leaves a device. That is the pitch for law, finance, healthcare, and other regulated shops that want agentic tools without dumping client records into someone else’s cluster.

Competitors are chasing the same mix of on-device privacy and cloud intelligence. Perplexity’s claim is that it can start a job in the cloud, peel off the confidential pieces mid-task, and finish without losing context. For Mac users sitting on 24GB-plus Apple silicon — and for the new M5/M6 machines Apple just put on the table — that is now a shipping product rather than a Computex slide.

MacDailyNews Take: The use cases Perplexity and early reviewers walk through are familiar: an attorney updating a brief with client files that stay on the Mac while case-law research runs in the cloud; an analyst rebuilding a confidential model against public comps; a small-business owner starting pricing research from a phone while private price lists never leave the desktop. In each case the agent is only as useful as the data it can touch. Hybrid Compute is an attempt to let it touch private data without treating the cloud as the default home for that data.

---

‎
Please help support MacDailyNews — and enjoy subscriber-only articles, comments, chat, and more — by subscribing to our Substack: macdailynews.substack.com. Thank you!

Support MacDailyNews at no extra cost to you by using this link to shop at Amazon.

The post Perplexity brings hybrid compute to Mac, keeping sensitive AI work on device appeared first on MacDailyNews.

Earlier today, Apple unveiled the new generation of Mac mini and Mac Studio, featuring the latest entries in the Apple silicon family of chips: the M6, available in the Mac mini, and the M5 Ultra, exclusive to the Mac Studio. You can read more details about the announcement and related specs in John’s overview.

As someone who’s been working with an M3 Ultra Mac Studio (on loan from Apple) with 512 GB of RAM for the past year (plus two separate M4 Mac minis) with a particular focus on local AI models and performance gains enabled by Apple’s MLX framework, I obviously am very interested in these new machines. To give you some context: my current workspace for my upcoming iOS and iPadOS 27 review, which lives in Notion, is entirely managed by a series of local agents running the latest DeepSeek-V4-Flash via MLX on the Mac Studio, which continuously scan the project for new notes, sources, and research material that is automatically categorized and linked in the chapters that I’m writing. So, yes, I’m sure I’ll have more thoughts on the new Mac minis and Mac Studios soon. In the meantime, I thought it’d be fun to break down the local AI-related details from Apple.

First, the M6. While it may not be the powerhouse that the M5 Ultra is, it’s an interesting new chip because of its new 2 nm process and dual Neural Engine system:

M6 is built using cutting-edge 2 nm process technology, packing greater transistor density into a smaller die for a major leap in performance and power efficiency. M6 also introduces a Dual 16-core Neural Engine, providing up to 2x the peak compute over previous generations to make on-device AI workflows run even faster. System frameworks can automatically utilize both engines simultaneously, enabling applications to see faster model execution.

But that’s not all: the new 12-core GPU (which has two more cores than the M4’s old GPU) now features Neural Accelerators in each core for the first time on the Mac mini. The result, according to Apple, is “4x faster AI performance and 2x faster graphics than Mac mini with M4”. If these numbers hold up and scale linearly, a local Mixture-of-Experts model such as Qwen 3.5-35B-A3B, which would run at ~17 tokens/sec on average on a base M4 Mac mini with 16 GB of RAM, could realistically generate output at over 60 tokens/second with the base model M6 Mac mini.

Based on what we’ve seen so far, the one downside of the M6 Mac mini is that it does not have Thunderbolt 5 ports; those are exclusive to the M5 Pro model, also announced today. As Apple also notes, Thunderbolt 5 allows multiple M5 Pro Mac minis to run local AI models as a single cluster (for example, using something like exo). Alas, that won’t be possible with the M6 Mac mini and its Thunderbolt 4 bandwidth bottleneck. Nonetheless, I’d be intrigued to compare the performance of the widely regarded Qwen3.8-27B and DeepSeek-V4-Flash models on the new M6 Mac mini.

Moving on to the flagship M5 Ultra Mac Studio, let’s start here:

With M5 Ultra, Mac Studio achieves up to 4.3x the peak AI compute performance of M3 Ultra and a staggering 9.8x more than M1 Ultra. Combined with up to 512GB of unified memory and 1.2TB/s of memory bandwidth, 50 percent higher than before, Mac Studio lets users run massive models entirely on device with complete privacy — without counting tokens or worrying about rising cloud costs.

As seen on the Mac Studio’s product page, the 512 GB RAM version of the M5 Ultra Mac Studio will only be available in “late October”. Unlike the M6, the M5 Ultra supports Thunderbolt 5, which will allow customers to run multiple Mac Studios in a cluster with shared memory to distribute inference across machines. And, according to Apple, a cluster of four Mac Studios yields three times the performance of a single Mac Studio.

Speaking from personal experience, I know that my M3 Ultra Mac Studio using oMLX can run DeepSeek-V4-Flash locally with generation averaging 35 tokens/second. Assuming a linear 4x increase, that would put the same model at over 120 tokens/second on an M5 Ultra Mac Studio. To put things in perspective, that kind of performance would be faster than any AI chatbot website, it’d be faster than many providers who offer a “fast” mode for their models, and it’d only be second to either dedicated NVIDIA PC clusters at home or specialized inference providers such as Cerebras or Groq…which are running in full-blown data centers. Sure, you would need a computer that is likely going to cost more than $20,000 to make it happen, but it’d still be possible on a single machine that is small, quiet, and that – in theory – any consumer can buy off the shelf.

It’s also interesting to consider the implications of M5 Ultra for running much larger models that would have typically required aggressive quantization to even stay in memory and produce any useful output. For example, Z.ai’s GLM-5.2 (a 744B MoE model) could run at up to 17 tokens/second on an M3 Ultra Mac Studio with 512 GB of RAM; the same person was even able to get the massive Kimi K3 model to run on a single machine…with decoding performance up to 3 tokens/second. You see where this is going: by continuing to push the envelope of what is possible with M-series chips, Apple is gradually making it possible to run frontier open-weight models on local hardware that takes up as much space as four Mac minis stacked together.

Obviously, for now, this is all theoretical: we’re going to need actual benchmarks to objectively measure the improvements and performance gains of the new chips. But if Apple’s claims hold up – and I have no reason to believe they won’t – I believe both the M6 and M5 Ultra will meaningfully shake up the landscape of local AI models on macOS, reinforcing, once again, that Apple continues making the best computers for AI in the industry.