The Legend of Zelda: Tears of the Kingdom, um dos jogos para Nintendo Switch e Nintendo Switch 2 mais populares e aclamados dos últimos anos, ganhou uma versão recompilada que pode ser rodada nativamente em Macs com Apple Silicon.
Mais um fruto do projeto MacNativeGames, o port roda em resolução 1440p e com ótimo desempenho no macOS, chegando a 90 quadros por segundo.
A versão usa a API 1Application programming interface, ou interface de programação de aplicações. Metal 4 da Apple e o MetalFX — tecnologias que minimizam os engasgos e deixam a jogatina mais fluida. Apesar disso, segundo o Notebookcheck, explorar novas áreas do mundo aberto de Hyrule pode causar travamentos momentâneos.
O grande desafio é o processo de configuração do jogo, já que os próprios usuários precisam fornecer uma ROM dele nos formatos XCI ou NSP — além de rodar uma série de comandos usando as ferramentas de desenvolvimento Xcode.
Esse processo de instalação trabalhoso do port, cujos arquivos podem ser baixados gratuitamente do GitHub, existe para evitar possíveis problemas com a Nintendo envolvendo direitos autorais — que, como sabemos, é bem chatinha nesse quesito.
It happens every time: Just before major OS releases, the tech web is flooded with how-to and tips articles. The MacVoices Live! panel looks at why and how you can benefit, including how that flood can offer guidance when new operating systems arrive. The discussion then turns to the Mac App Store allowing developers to […]
Apple is working on an AI server that would use Apple’s high-performance M-series Ultra chips found in Mac desktops. The potential product’s expected release in 2029 would mark the first Apple server to hit the market in nearly two decades—and could capitalize on the surging popularity of Apple hardware among AI developers.
The enterprise server would come in two configurations that include either two or four of Apple’s future M8 Ultra chips, according to The Information. The project reportedly received support from new Apple CEO John Ternus when it began a year ago, back when Ternus led Apple’s hardware engineering efforts.
This revelation coincides with booming sales for Apple’s Mac mini and Mac Studio as AI developers and companies snap up the Mac computers to run AI workloads. The popularity of such computing devices that rely on Apple’s M-series Ultra chips has undoubtedly encouraged Apple’s pursuit of an enterprise server using the same chips.
Apple Intelligence is back, baby! That's right: Two years after its initial launch, Apple is giving its generative AI features their first significant upgrade, along with a new version of Siri that the company has been strongly insinuating would be a part of each of the last two macOS releases, without ever actually announcing a release date.
Generative AI is unavoidable if you install macOS 27 Golden Gate. I'm being very literal: There used to be a toggle you could hit if you wanted to turn off Apple Intelligence and delete the gigabytes' worth of AI models it would download to your disk, and now there isn't. Apple Intelligence defines the Golden Gate release, in everything from its marketing to its features to the chips it runs on to the amount of space it takes on your disk. Ready or not, here it is.
But if you can manage to look past that fanfare somehow, the operating system underneath Apple Intelligence is getting the kind of update that macOS users usually proclaim to like. Golden Gate addresses the most irritating of macOS 26 Tahoe's design sins, adds a big pile of incremental improvements, and promises a bunch of under-the-hood fit-and-finish optimizations to make common tasks feel just a little faster and more reliable than before.
Apple Intelligence is back, baby! That's right: Two years after its initial launch, Apple is giving its generative AI features their first significant upgrade, along with a new version of Siri that the company has been strongly insinuating would be a part of each of the last two macOS releases, without ever actually announcing a release date.
Generative AI is unavoidable if you install macOS 27 Golden Gate. I'm being very literal: There used to be a toggle you could hit if you wanted to turn off Apple Intelligence and delete the gigabytes' worth of AI models it would download to your disk, and now there isn't. Apple Intelligence defines the Golden Gate release, in everything from its marketing to its features to the chips it runs on to the amount of space it takes on your disk. Ready or not, here it is.
But if you can manage to look past that fanfare somehow, the operating system underneath Apple Intelligence is getting the kind of update that macOS users usually proclaim to like. Golden Gate addresses the most irritating of macOS 26 Tahoe's design sins, adds a big pile of incremental improvements, and promises a bunch of under-the-hood fit-and-finish optimizations to make common tasks feel just a little faster and more reliable than before.
An unverified Geekbench 7 result attributed to Apple’s M6 chip has appeared ahead of the new Mac mini’s September 22 release, offering an early and incomplete look at potential CPU gains. Geekbench is a synthetic benchmark designed to compare processor performance through a defined workload. Its result can be useful as a point of comparison, […]
The first unverified Geekbench 7 result for Apple’s M5 Ultra suggests a substantial CPU gain over M3 Ultra, but the single database entry is not a substitute for independent testing. The listed result is for the highest-end M5 Ultra with a 36-core CPU. It should not be assumed to represent every Mac Studio configuration or […]
Resultadospreliminares de benchmark do chip A20 Pro, que alimenta o iPhone 18 Pro, o iPhone 18 Pro Max e o iPhone Duo, surgiram na manhã de hoje no banco de dados do Geekbench 7, oferecendo uma primeira visão do desempenho de CPU1Central processing unit, ou unidade central de processamento. e GPU2Graphics processing unit, ou unidade de processamento gráfico. dos novos modelos antes mesmo de chegarem ao mercado.
Com base nesses resultados iniciais, o chip A20 Pro parece oferecer um desempenho cerca de 21% a 23% mais rápido em single-core (média de 4.722 pontos), além de 27% a 28% mais rápido em multi-core (média de 12.626 pontos) em comparação com o chip A19 Pro dos iPhones 17 Pro/17 Pro Max.
Talvez não surpreenda que os resultados do iPhone 18 Pro Max sejam praticamente iguais aos do iPhone 18 Pro. Uma diferença aparece nas pontuações do teste Metal: o modelo menor marcou 64.069 pontos, enquanto o maior atingiu 64.915 pontos.
Ainda assim, os resultados gráficos alinham-se, em linhas gerais, à afirmação da Apple — presente no comunicado de imprensa e na página do iPhone 18 Pro — de que o A20 Pro oferece uma performance gráfica até 40% mais rápida que o chip A19 Pro.
Também foram vistos resultados de benchmark do iPhone 18 Pro Max no AnTuTu, que registrou uma pontuação geral de 3.551.073. A pontuação da CPU, especificamente, foi de 920.410, enquanto a GPU alcançou 1.397.374 pontos. Os testes de memória e UX registraram 470.245 e 763.044 pontos, respectivamente. Para fins de comparação, o iPhone 17 Pro Max obteve uma média de 2.275.118 pontos, colocando o novo modelo substancialmente à frente no benchmark.
Vale ressaltar que esses testes de desempenho pontuais — que ainda não foram confirmados — não indicam como os dispositivos se comportarão durante sessões prolongadas de jogos ou outras cargas de trabalho exigentes; afinal, os resultados de testes no mundo real poderão apresentar algumas diferenças.
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.
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!
Apple’s new Mac mini and Mac Studio spark a discussion of the M6 and M5 Ultra chips and how they measure up in performance and upgrade decisions. The panel weighs Mac Studio trade-ins, Mac mini value, on-device AI and large language models, and Apple’s increasingly powerful silicon. They also examine the move to 2-nanometer technology and debate Apple’s redesigned keyboards and glyph-only modifier keys.
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.
We got a nice surprise from Apple this morning with some new Macs. Those of you waiting for that M5 Ultra Mac Studio are finally going to get your wish. You can get the M5 Ultra with up to 36 CPU cores and up to 80 GPU cores. Interestingly, memory tops out at 512 GB. That’s the same ceiling the M3 Ultra had, and it’s the configuration Apple dropped back in March. There’s no mystery why. The M5 Ultra starts at $5,500, up from the $4,000 entry price for the M3 when it launched. Again, no surprise.
If your needs are more modest, the Mac mini also got updated to M6 and M5 Pro variants. That’s right, we can now buy an M6 Mac.
The mini also took a price hike. The base model is $899 now, up from $599 last fall and $699 since June. I’m glad Apple released these things. I expect the AI gang will buy them like hotcakes.
At NAB 2026, Russell Pompea, Account Executive from Topaz Labs, discusses major advances in Topaz Video, including Starlight Precise 2.5 for restoring low-quality footage, SDR-to-HDR conversion, and Apple Silicon optimization. The conversation also covers Topaz’s photography tools, model customization, subscription pricing, local vs. cloud processing, and shared responsibility for cloud security.
Há dois meses, falamos aqui sobre o RetinaDesk, um site muito bacana que ajuda você a escolher o melhor monitor externo para o seu Mac, levando em conta todas as especificidades dos computadores da Apple.
Pois bem, recentemente, a ferramenta ficou ainda mais completa com a adição de uma página que ajuda você a descobrir, de maneira extremamente intuitiva, a quantidade máxima de displays externos suportados pela sua máquina.
O banco de dados da ferramenta cobre todos os Macs com Apple Silicon lançados até hoje, bem como suas respectivas variantes — detalhe que joga o total de computadores listados pelo RetinaDesk para nada menos que 53.
Encontrar essa página é fácil: basta acessar o site do RetinaDesk, clicar na aba “Tools” na parte de cima da tela e, depois, selecionar a opção “Mac External Display Support by Chip” — ou simplesmente acessar esse link. Feito isso, é só selecionar o modelo do seu Mac na seção “Which Mac is it?” e, em seguida, sua respectiva variante — indicada pelo chip e o ano de lançamento (e, às vezes, o número de portas do computador) — em “Step 2”.
Captura de tela/RetinaDesk
Uma vez que você indicar o modelo do seu Mac, o número máximo de monitores externos ao qual ele pode ser conectado deverá aparecer à direita da tela, junto de uma ilustração do seu “setup máximo” — ou seja, com o maior número de monitores possível.
A nova página, no entanto, não para por aí, já que ela também conta com outra seção a qual ajuda você a montar setups com diferentes combinações de monitores, os quais podem ter resoluções e taxas de atualização distintas — sempre levando em conta, é claro, as limitações do seu computador. Ela também é uma boa forma de descobrir se o seu Mac suporta uma resolução ou uma taxa de atualização (ou uma combinação dessas duas coisas) específica.
Captura de tela/RetinaDesk
A seção “What each built-in port can drive”, por sua vez, exibe com quais resoluções/taxas de atualização cada porta do seu computador é compatível, além de revelar a sua versão (se é Thunderbolt 4 ou 5, por exemplo). Logo abaixo, temos a seção “How you can arrange the displays” que, como seu nome indica, mostra todas as combinações possíveis de monitores para a sua máquina.
Por fim, a ferramenta também deixa você comparar um Mac qualquer com outro modelo da sua escolha, checar se sua máquina precisa estar no modo clamshell para suportar um setup específico e aprender sobre soluções alternativas e suas respectivas concessões.
Captura de tela/RetinaDesk
Assim como o resto das ferramentas do RetinaDesk, essa nova página é totalmente gratuita, com o único “porém” sendo o fato de que ela não está disponível em português — problema que, é claro, pode ser facilmente remediado com a ajuda das ferramentas de tradução dos navegadores atuais.
O MacBook Air foi atualizado neste ano com o chip M5 e é o notebook ideal para quem quer um Mac portátil capaz de lidar com a grande maioria das tarefas diárias. Além disso, ele traz um visual moderno, com uma tela Liquid Retina e o conector MagSafe 3.
Se você está de olho nesse notebook, estamos trazendo hoje duas ofertas muito boas no KaBuM! — atrativas principalmente levando em conta os recentes aumentos de preços na linha de computadores da Maçã.
Modelo de 13 polegadas
O modelo de 13″, na cor prateada, traz 16GB de memória unificada, 1TB de SSD1Solid-state drive, ou unidade de estado sólido. e chip M5 com uma GPU 2Graphics processing unit, ou unidade de processamento gráfico. de 10 núcleos.
MacBook Air de 13” (M5 com GPU de 10 núcleos, 16GB de memória e 1TB de SSD)
Parcelado: de R$19.599,00 por R$12.902,10 em até 10x À vista no Pix: de R$17.639,10 por R$12.257,00
Ele está saindo por R$12.902,10, valor que pode ser parcelado em até 10 vezes no cartão de crédito. Se levarmos em conta que o preço tabelado desse mesmo computador é de R$19.599, estamos diante de um desconto de 34%.
Para quem preferir pagar à vista no Pix, o preço cai para R$12.257,00, o que representa um desconto de 31% se compararmos com o valor à vista praticado pela Apple, de R$17.639,10.
Para pagamentos à vista no Pix, o preço fica melhor: cai para R$13.667,99. Como o valor à vista praticado pela Maçã no país é de R$19.439,10, estamos diante de um desconto de 30%.
Para aproveitar ambas as ofertas, aplique o cupom 8DO8 no carrinho.
Todos os descontos de promoções divulgadas pelo MacMagazine são calculados com base nos preços sugeridos pela Apple ou por outras fabricantes. Pode ser que um determinado produto seja comumente encontrado por valores inferiores em redes varejistas, mas a nossa base de comparação é sempre em cima das tabelas oficiais.
Não perca! 💻
Fique sempre por dentro das ofertas!
Se você não é louco, é claro que gosta de economizar. E, para isso, o MacMagazine tem diversas opções para ajudar você a conseguir comprar o seu produto Apple pagando menos!
NOTA DE TRANSPARÊNCIA: O MacMagazine recebe uma pequena comissão sobre vendas concluídas por meio de links deste post, mas você, como consumidor, não paga nada mais pelos produtos/serviços usando os nossos links de afiliado.
Notas de rodapé
1
Solid-state drive, ou unidade de estado sólido.
2
Graphics processing unit, ou unidade de processamento gráfico.
Apple Intelligence files can occupy several gigabytes of Mac storage because locally downloaded models power text generation, summaries, image creation, Siri requests, and other on-device features. Removing these files can recover valuable space when Apple Intelligence is no longer needed.
Apple lists 7GB of available device storage among the requirements for Apple Intelligence, but the amount displayed on a Mac can be higher. Additional languages, image-generation resources, updated models, and temporary assets can increase the total over time.
The files are managed by macOS and do not appear as ordinary documents that can be moved to the Trash. Apple provides a supported removal method through System Settings: turning off Apple Intelligence causes macOS to remove the downloaded on-device models.
This also disables features that depend on Apple Intelligence. Writing Tools, summaries, Image Playground, Genmoji, intelligent Shortcuts actions, and Apple Intelligence functions inside Siri may become unavailable. Regular Mac applications and the rest of macOS continue working normally.
Check Apple Intelligence Files on Your Mac
Recording the available storage before removing the files makes it easier to confirm how much space macOS recovered afterward.
To see the current storage allocation:
Apple menu > System Settings > General > Storage
Allow the Storage page to finish calculating its categories. The process may take a moment when the Mac contains many applications, local files, connected volumes, or several user accounts.
Apple Intelligence may appear as a separate item within the macOS storage category. Depending on the installed macOS version, some related assets may instead be counted under macOS or System Data.
Note the amount of available storage displayed near the storage graph. Comparing the total free space before and after the procedure is usually more reliable than watching one category because macOS can reclassify system-managed files while deleting them.
To inspect the complete internal storage container:
Apple menu > System Settings > General > Storage > All Volumes
This view is useful when the internal drive contains several APFS volumes sharing the same physical storage. The available capacity belongs to the shared container rather than remaining permanently assigned to one volume.
The storage used by Apple Intelligence is different from the memory, or RAM, used while features are running. Closing applications may reduce RAM consumption, but it does not delete the models stored on the internal SSD.
Image Credit: AppleMagazine
Remove Local Apple Intelligence Files
Apple Intelligence is available on compatible Macs with Apple silicon. Intel-based Macs do not support the complete Apple Intelligence system and should not contain the same downloadable model collection.
To turn off Apple Intelligence and request the removal of its on-device models:
Apple menu > System Settings > Apple Intelligence & Siri > Turn off Apple Intelligence > Turn Off Apple Intelligence
The second Turn Off Apple Intelligence command confirms the change. The exact appearance may vary between macOS versions, with the first control presented as either a switch or a button.
Apple says that turning off Apple Intelligence removes the on-device models. There is no separate model manager, Delete Models button, or supported Finder location from which users must remove the files manually.
Keep the Mac connected to power while macOS completes its background storage work. The Apple Intelligence setting changes immediately, but the Storage page may not display the recovered capacity at once.
macOS manages these models through its system asset service. The operating system must identify the associated packages, release them, and recalculate the storage categories. The process can therefore continue after the Apple Intelligence switch has already moved to the off position.
Turning off only individual features does not remove the complete model collection. Disabling notification summaries, Writing Tools, ChatGPT access, or Siri suggestions leaves Apple Intelligence itself active.
A Screen Time restriction also serves a different purpose. It can prevent access to image generation, Writing Tools, and other functions without turning off the underlying intelligence system. Use the Apple Intelligence control inside Apple Intelligence & Siri when the objective is recovering storage.
Confirm That the Space Was Recovered
After macOS has had time to process the removal, return to the Storage page.
To check the available capacity again:
Apple menu > System Settings > General > Storage
Wait for the storage graph and categories to finish updating. Compare the new amount of available space with the figure recorded before Apple Intelligence was turned off.
The Apple Intelligence entry may disappear completely, display a smaller amount, or become part of an adjusted macOS or System Data category. The increase in total available storage provides the better indication that the local models were removed.
If the figures have not changed, restart the Mac to refresh its system services and storage calculation.
To restart macOS:
Apple menu > Restart > Restart
Sign back in, allow the Mac to remain idle for a while, and open the Storage page again.
Apple menu > System Settings > General > Storage
Restarting does not delete the models by itself. Apple Intelligence must remain disabled while macOS completes the removal.
Emptying the Trash is not part of this procedure because macOS does not place Apple Intelligence models there. The Trash affects only files that a user or application has already marked for deletion.
Image Credit: Apple Inc.
What to Do When the Files Remain
First, verify that the complete Apple Intelligence system is off rather than one related feature.
To confirm its status:
Apple menu > System Settings > Apple Intelligence & Siri > Apple Intelligence
The control should display Apple Intelligence as disabled. If macOS presents another confirmation request, complete it and leave the feature off.
Storage information can take time to settle because macOS also manages caches, local snapshots, temporary update packages, and purgeable files. These categories can change while the Mac is idle, making an immediate before-and-after comparison misleading.
Installing the latest stable macOS update can correct storage reporting or asset-management problems.
To check for an available update:
Apple menu > System Settings > General > Software Update
Save current work before installing an update. The Mac may need to restart, and the installation temporarily requires additional free storage.
Beta versions of macOS require extra caution. Development releases may download experimental models, retain previous model revisions, or temporarily change the Apple Intelligence controls. A procedure confirmed for the public version of macOS may not produce the same storage behavior in a beta.
Avoid following online instructions that require starting in macOS Recovery, disabling System Integrity Protection, mounting protected volumes, and deleting folders with names referring to MobileAsset or generative models.
System Integrity Protection prevents applications and administrator accounts from modifying protected parts of macOS. Disabling it weakens that protection and creates the possibility of deleting the wrong asset, leaving macOS with an incomplete model installation, or causing the system database to disagree with the files present on the drive.
A storage utility can help identify how much space system files occupy, but it should not require disabling macOS security to remove Apple Intelligence. If the documented switch leaves a large allocation after a restart and sufficient processing time, Apple Support can investigate the asset database without exposing the Mac to an unsupported system modification.
Turning Apple Intelligence on again reverses the storage saving.
To restore its features:
Apple menu > System Settings > Apple Intelligence & Siri > Turn On Apple Intelligence
The Mac will download the required local models again. Apple recommends leaving the device connected to Wi-Fi and power while the model download completes.
The speed of laserlight is due for a major upgrade after an international collaboration of scientists has developed a breakthrough nanomaterial. The new metasurface can steer and shape light beams at unprecedented speeds of just 74 quadrillionths of a second.
Scientists at UC Berkeley and Tel Aviv University revealed their new optical surface, developed in a Caltech lab, in a recent paper published in Nature Nanotechnology.
According to the team behind the breakthrough, their work will have ramifications for communications, sensing, imaging, and even quantum technologies, providing a new benchmark for “high-speed” light manipulation.
Light Control
Since the fiber-optic advancements of the 1980s, lasers and light pulses have underpinned some of our most crucial cutting-edge technologies. Advanced computers, communications systems, sensors, and even the elusive quantum systems all rely on controlling light.
While the speed of light is so incredibly fast that physicists consider it a constraint on the maximum speed that matter can travel through the vacuum of space, the speed at which our technology reacts to it is far slower.
The scientists’ breakthrough metasurface contains silicon structures even smaller than the wavelength of light, resulting in an ultra-thin surface. Silicon’s optical properties change upon bombardment with laser pulses, a property that the team strongly enhanced with their nanostructure design. In their design, the properties of the material are modified via light pulses, which alter how the material influences a second beam of light impacting its surface. Notably, their metasurface allows for control of a beam of light without any moving parts, such as mirrors or lenses.
Real World Results
The team conducted laboratory tests in which they manipulated light beams by up to 13 degrees and reshaped them into different patterns.
“The key idea was to create a metasurface whose optical response is not fixed once it is fabricated,” said lead author Dr. Claudio Hail. “By changing the illumination pattern, we can reconfigure how the device steers and shapes light, and do so on an ultrafast timescale.”
“For me, the most exciting aspect is seeing an idea that had been with us for years become an experimental reality,” said co-author Dr. Lior Michaeli. “The ultrafast effect we wanted to use is naturally very weak. We had to design the metasurface so that it would amplify the effect enough not only to measure it, but to use it to steer and shape light.”
While the metasurface produced record-breaking speeds under laboratory observations, exactly how fast it can operate remains uncertain. The researchers note that the switching time and the duration of the laser pulse were extremely close, which indicates that quicker pulses could push the optical surface’s speed even higher.
The Future of Light Technologies
One of the major potential advances that could come from this new metasurface includes systems that carry information primarily with light, end-to-end. Current communications systems use light to transmit messages, but those messages are converted into electrical signals for processing. By increasing the speed at which light can be manipulated, this would reduce the necessity to convert light signals into electrical ones, removing unnecessary actions in computing.
“This work points to a broader opportunity: using engineered structures to strengthen interactions between light and matter and turn them into tools for control, sensing, and information processing,” Dr. Michaeli said. “This is closely connected to research directions we are now pursuing at Tel Aviv University.”
Despite its promise, the work is still in the fundamental phase. Although not yet ready for practical use, the metasurface offers promise in future ultra-fast computing applications, where it may soon help to cut out the middleman of electronic conversion.
Ryan Whalen covers science and technology for The Debrief. He holds an MA in History and a Master of Library and Information Science with a certificate in Data Science. He can be contacted at ryan@thedebrief.org.
The reckoning is coming. Apple has done it before, and it’s doing it again. After macOS 27 Golden Gate, you will no longer be able to run legacy Intel apps using Rosetta 2. After the 26.4 update in March, when running any affected app in macOS 26 Tahoe, you see a notification banner when you launch affected software or a background component launches that uses Intel code:
Support Ending for Intel-based Apps. This version of “[App Name Here]” will not open in a future release of macOS. Learn how to update to an Apple silicon version. [Learn More pop-up menu]
Whenever you launch an Intel-only app, macOS politely tells you the end is coming.
There is one carveout noted for “certain older, unmaintained games that rely on Intel-based frameworks,” so far unnamed.
With information in hand, you can plan how to deal with Intel-only apps that have been abandoned, or the developers have seemingly no interest in updating. Printer makers are, so far, the worst.
A history of looking backward
Apple has done this before—many times, in fact. I’ve argued in the past that one of Apple’s key strengths, and a key to the continuity of the Mac’s operating system, is how well it’s managed transitions across chip architectures, frameworks, and operating system changes.1
Look at what Apple has managed:
Motorola 680×0 to PowerPC (silent emulator)
Mac OS 9.1 (or later) to Mac OS X (Classic)
PowerPC to Intel (Rosetta, original flavor)
32-bit to 64-bit code (warnings)
Intel to Apple silicon M-series chips (Rosetta 2)
At each of those milestones, Apple offered either advance warnings about a future lack of compatibility, an emulator tool, or both. Does anyone have fond memories of running QuarkXPress using Classic in Mac OS X in its early days? When it crashed, which was often, it only took down the Classic “app” and didn’t require a Mac restart.
To be fair, Apple could have continued to let older software run for as long as it was willing to devote time to maintaining emulators or compatibility layers. We could be using apps written for Motorola 68030s on our M5 Macs right now! However, it makes increasingly less sense to support the past after a reasonable transition, particularly one that stretches years and sees the company discontinuing compatible hardware.
Apple released the last Intel-based model in August 2020, introduced its first M-series computer in November 2020, and halted sales of computers with Intel inside in 2023.
As part of this shift, you receive the Intel software warning, not just the first time you launch an affected app or an Intel-based component or even Terminal-accessed binary runs, but the next time you do it after a restart. Yes, it is trying to bug you to get a newer version of your software. Golden Gate also pulls an interesting trick: When you upgrade, it removes the Rosetta emulator. You’re prompted to install it the first time you need Intel code. I guess this is an extra spur to remind you.
Because Intel-only apps will continue to work under Rosetta until you install macOS 28 Dumbarton, these reminders will continue for at least another 16 months or so.2 Once you upgrade to macOS 28, those apps and components will simply stop working.
Sunsets can be beautiful
Use Golden Gate to review a list of Intel-only apps, although third-party apps can provide more information.
Apple includes a helpful tool in Golden Gate in System Settings: General: About. Click Details next to Intel-based Apps, and a dialog box appears that lists everything installed that requires Rosetta. As a nice touch, Apple shows you apps launched in the last year, along with when, so you know the most critical to upgrade or swap out.
Golden Gate will list all the apps (but not all components) that require Rosetta.
You can also examine any app in the Finder, choose Get Info from the File menu, and note the warning next to Kind. Apple allows apps to be compiled as universal binaries that include both Intel and Apple silicon code, and offers an Open using Rosetta checkbox that was useful in the early days of the swapover when some M-series versions of an app were missing features or were buggy compared to the Intel version. You sometimes checked that box to use the more stable Intel variety, and then surely forgot about it. (For instance, Adobe updated its primary apps faster than third parties did their plug-ins. To use some Photoshop plug-ins, you had to force Photoshop to run in emulation.)
The Get Info dialog is a way to check on the code status of an app.
Hover over the item, and you can click Visit Website if Apple has a link to one, or click the More … button and choose Show App in Finder or Move to Trash. Not every app or component has a Move to Trash item, and I’m not sure why. I also suggest installing AppCleaner, a free app that will root out all files associated with an app when it’s deleted. (Turn on SmartDelete to monitor the Trash, or you can drag apps onto it.) This ensures you’re not leaving cruft behind when you get rid of the app proper.
This is all fine for rooting out installed apps, but Apple may not catch every system component, command-line utility, or other software that lives in crevices and emerges in particular conditions. For that, you should take a look at:
Go64: The free app from St. Clair Software, the makers of our favorite app, Default Folder X, provides details in a list form.
Rosetta Check ($7): Turn to Rosetta Check for even more information about each unsupported app, which may be useful if you have a lot of software to work through updating.
HoudahSpot: If you use HoudahSpot, you can construct a search.
A few companies that have Intel-only apps or components have explained their migration path, with PaperCut being an excellent example. Others without any financial incentive—such as hardware makers, who sometimes barely care about compatibility for software required to support their products—may wait until after the 2027 transition to wake up and make changes. This has happened many times before!
[Got a question for the column? You can email glenn@sixcolors.com or use/glennin our subscriber-only Discord community.]
Login
