logo
Inside Apple's new AI models: How it works, where it gets its training data

Inside Apple's new AI models: How it works, where it gets its training data

Apple unveiled upgraded AI models designed to power Apple Intelligence features across iPhones, iPads, and Macs at WWDC 2025. These include both on-device and cloud-based models. Now, Apple has published a detailed technical report titled "Apple Intelligence Foundation Language Models Tech Report 2025", outlining how these models work and where their training data comes from. Here's a breakdown of the key points:
What are Apple's new AI models
Apple introduced two foundation models that support Apple Intelligence across apps and services. The first is a compact 3-billion-parameter model designed to run directly on Apple devices powered by Apple Silicon. The second is a more powerful server-based model hosted on Apple's Private Cloud Compute (PCC) infrastructure.
Both models are multilingual and multimodal, meaning they understand multiple languages and can process different types of content, including text and images. Apple said that they were trained using large-scale datasets sourced through licensed content, web crawling, and synthetic data generation.
How do these new AI models work?
On-device model:
Starting with the on-device AI model, Apple split this model into two parts to reduce memory usage and make it faster:
Block 1: Handles most of the processing.
Block 2: Skips some operations to save memory and boost speed.
This architecture helps the model respond faster without losing quality, making it more suitable for real-time features like text suggestions or summarisation.
Cloud-based model:
The cloud-based model uses a more advanced architecture called Parallel-Track Mixture-of-Experts (PT-MoE). Instead of processing every task with the full model, PT-MoE routes each task to the most relevant "experts"—specialised mini-models trained for specific content types. So, for example, if you ask it to plan a vacation, only the travel-related experts get activated.
This not only speeds things up but also makes the model more efficient. Apple also created a new kind of Transformer (a type of neural network) that processes multiple parts of a request in parallel, reducing bottlenecks and improving performance.
What are the benefits of these new AI models?
One of the biggest improvements is expanded multilingual support. Apple increased the portion of training data in non-English languages from 8 per cent to 30 per cent. It also expanded the model's vocabulary from 100,000 to 150,000 tokens.
This means Apple Intelligence can now better understand and respond in more languages, with improved fluency and accuracy. Apple said that it has also tested this using prompts written by native speakers, ensuring the models performed well across different cultures and languages. With this, features like Writing Tools should now work more reliably outside of English.
Apple has also opened access to its on-device model for third-party developers, allowing them to use AI-powered features like summarisation or rewriting directly within their apps, without sending data off the device.
From where did Apple source its data?
Apple said that it trained its AI models using a wide range of high-quality data, but it does not use your private information or personal device activity.
Instead, Apple said it relied on the following main sources:
Licensed content from publishers
Publicly available and open-source data
Web content collected by Applebot, Apple's own web crawler
Apple said that its web crawler respects 'robots.txt' rules that let websites opt out of being used for training. Publishers can also limit which pages are accessible while still being included in Apple services like Siri and Spotlight search.
Text data:
A big part of Apple's training data came from websites. The company said that Applebot crawled billions of web pages across different topics and languages and used smart tools to load full pages, interact with dynamic content, and extract useful information. To ensure quality, Apple used AI-based filtering instead of rigid rules, which helped retain more relevant content while avoiding low-quality or inappropriate data.
Image data:
To help its models understand visuals, Apple trained them on image-text pairs. It used licensed images, public images with captions, and AI-generated image descriptions. Apple also included visual materials like infographics, tables, and charts. For example, it used AI to create sample data, generate charts, and then produce related questions for training.
Orange background

Try Our AI Features

Explore what Daily8 AI can do for you:

Comments

No comments yet...

Related Articles

Apple iPhone 17, iPhone 17 Air, iPhone 17 Pro, iPhone 17 Pro Max shocking new looks, colours revealed? Details here
Apple iPhone 17, iPhone 17 Air, iPhone 17 Pro, iPhone 17 Pro Max shocking new looks, colours revealed? Details here

Time of India

timean hour ago

  • Time of India

Apple iPhone 17, iPhone 17 Air, iPhone 17 Pro, iPhone 17 Pro Max shocking new looks, colours revealed? Details here

Apple has rumoured line up of a number of iPhones. Apple iPhone 17 , iPhone 17 Air , iPhone 17 Pro, and iPhone 17 Pro Max are likely to be released in a number of colours, according to reports citing alleged leaks. This comes as Bloomberg News reported that Apple was planning new Macbook pro, iPhone 17e and iPads by early 2026. Apple has introduced Applecare one, streamlining coverage into a single plan with incredible value. Apple Inc stated that Applecare one costs $19.99/month for three products, as per Reuters report. Apple iPhone 17 Explore courses from Top Institutes in Please select course: Select a Course Category PGDM Design Thinking Finance Data Science Digital Marketing Healthcare Cybersecurity Artificial Intelligence MCA Others Management healthcare Data Science CXO Data Analytics Public Policy Project Management Product Management Leadership MBA others Operations Management Degree Technology Skills you'll gain: Financial Analysis & Decision Making Quantitative & Analytical Skills Organizational Management & Leadership Innovation & Entrepreneurship Duration: 24 Months IMI Delhi Post Graduate Diploma in Management (Online) Starts on Sep 1, 2024 Get Details According to leaks, Apple iPhone 17 will come in multiple colours, including black and white. The other colours are reportedly steel gray, black, white, green, purple, and light blue. by Taboola by Taboola Sponsored Links Sponsored Links Promoted Links Promoted Links You May Like Join new Free to Play WWII MMO War Thunder War Thunder Play Now Undo Apple iPhone 17 Air Apple iPhone 17 Air will come in light blue. iPhone 17 Air will have other colour varieties such as black, white, light blue, light gold, as per the leaks. Apple iPhone 17 Air Pro, Apple iPhone 17 Air Pro Max Live Events Apple iPhone 17 Air Pro and Pro Max are likely to come in orange hue, black, white, gray, dark blue, as per leaks. Readers please note that all these details about Apple iPhone 17 variants are based on reported leaks and Apple is yet to make any confirmation. Meanwhile, at its Worldwide Developers Conference (WWDC) in June, Apple announced iOS 26, the next version of its iPhone operating system. Among the new features, there is a complete redesign with Apple's new "Liquid Glass" design language along with updates to apps across the platform, according to a report by The Verge. The announcement also marked a change in how Apple signifies its major updates to iOS. Under the previous marketing scheme, this year's major release would have been iOS 19 -- the direct follow-up to iOS 18. However, as per the new announcements at the WWDC 2025, Apple's big iOS updates will now be numbered based on the year following their introduction, reported The Verge. FAQs Q1. When did Apple WWDC happen? A1. Apple Worldwide Developers Conference (WWDC) held in June. Q2. How many iPhones Apple are planning? A2. Apple has rumoured line up of a number of iPhones. Apple iPhone 17, iPhone 17 Air, iPhone 17 Pro, and iPhone 17 Pro Max.

No hiring in India: Could Trump's new push disrupt tech jobs?
No hiring in India: Could Trump's new push disrupt tech jobs?

India Today

time2 hours ago

  • India Today

No hiring in India: Could Trump's new push disrupt tech jobs?

Donald Trump has called on Google, Apple, and other US tech giants to stop hiring in India and prioritise Americans instead. The statement, though not yet a policy, has triggered concern across India's white-collar job sector, especially among engineering and management graduates aiming for roles in global tech enforced, the impact could be sharp: from IIT placement trends to mid-level tech hiring, from GCC operations to India's growing dependence on global companies for high-skilled and Wisdom Hatch Founder Akshat Shrivastava said on X that the chances for Indian jobseekers are LET'S UNDERSTAND HOW THE HIRING WORKS American tech companies don't just hire from India. Some also build in India, including the Google, Apple, and Microsoft, all have large operations here, often in the form of Global Capability Centres (GCCs).These giants maintain a presence in India where professionals are at the task of building products, writing code, testing machines, overseeing global employs around 10,000 people in India. Microsoft has over 18,000 employees here. These are for operations across the has around 5,000 direct employees in India, and thousands more in its supplier network and development of these jobs are not for call centres or support. The core development roles, white-collar positions, are in play that attract graduates from IITs, NITs, IIITs, and even Tier-2 colleges.A pronounced number of these roles are built for India but are part of global projects. And yes, some top-tier students are hired directly to go to the US exactly the pipe Trump wants to close."There's no denying that companies like Google and Apple have long symbolised the 'dream job' for many Indian students, and rightfully so. Indian talent has been a major contributor to the global tech revolution, not just as employees but as leaders," says Neelakantha Bhanu, Founder and CEO of Bhanzu, and title holder of 'World's Fastest Human Calculator'."However, if such hiring freezes become a reality, it will be a wake-up call, not in fear, but in perspective," he says."The world is changing, and so are opportunities. India today is not just a source of talent, but a builder of global products," Bhanu adds. WHAT HAPPENS TO IIT AND IIM PLACEMENTS?Every December, the buzz begins: placement season. But behind the success stories, there's a truth not often told -- many students don't land dream IITians, despite the brand tag, end up in jobs that pay Rs 8-10 lakh per annum or less. Not because they aren't brilliant. But because not everyone gets picked by Google, Microsoft, or a US -- based startup with a fancy from IITs from 2023-2024 show that even in top IITs, 20-25% of students were still unplaced at the end of the obtained through RTI requests filed by IIT Kanpur alumnus Dheeraj Singh shows that nearly 8,000 students, around 38% across 23 IITs, remain unplaced in the year IIMs, especially the older ones, place most students in India-based roles, consulting, banking, and management BIG PICTURE: INDIA'S WHITE-COLLAR WALL COULD CRACKTrump's statement comes at a time when India is producing more engineers than it can absorb. Private colleges, deemed universities, and even Tier-1 institutes are churning out thousands of tech graduates each year. But demand has 51.25% of graduates amongst the graduates in India are considered employable, highlighting persistent gaps in vocational training and skill development, as per the Economic Survey turns out nearly 15 lakh engineering graduates each year, yet only 10-15% among them find employment, as noted in a report by has kept things afloat over the past decade is the globalisation of Indian tech talent. US -- Mbased hiring, remote work, global team integration, and GCC expansion have created a top 10% of tech graduates, those who would go abroad or work on US -- facing roles from India, may have to compete in a shrinking domestic market. And this creates a domino effect down the ladder."If US tech companies stop hiring from India, it'll cost them more than us. India has long been their strongest talent pool, from engineers to CEOs. Some students may miss out on overseas roles, but fewer than 2% of IIT graduates go abroad now," says Nishant Chandra, Co-founder, Newton School."Most choose to stay and lead from India. This shift could actually benefit us by putting focus on skills over pedigree," he SMALLER STARTUPS FOLLOW SUIT?Possibly. If the bigger players hit pause, mid-sized companies may rethink their hiring plans too, especially those who build for US clients or rely on US venture capital. And in India's startup ecosystem, perception drives could delay hiring cycles, reduce internship opportunities, and force more candidates to settle for lower Group Founder Ankur Agarwal, a top executive search firm, sums it up: 'These Trump rules, if enforced strictly, will definitely impact placements in IITs as the top US companies recruit quite a bit from these campuses for their US -- based tech development. IIMs are usually used to hire for India roles only, so they will not be impacted.""The real impact will be felt by the GCCs, though, which have become an important recruiter for top quartile tech talent. However, the actual impact will depend on how strictly companies comply with this directive and whether it becomes formal policy, as the US still faces significant tech talent shortages that make complete elimination of overseas hiring challenging," he NEXT?Nobody knows if Trump's statement will become law. But it's already a signal."Our institutions, our ecosystems, and our ambitions are ready. And as someone who chose to stay and build here, I can say that there's never been a more exciting time to be in India. We're not just producing global talent anymore. We're producing global solutions," says time to prepare is now, not just with coding skills, but with adaptability, global exposure, and maybe even a Plan B that doesn't rely on a Silicon Valley zip code.- Ends advertisement

Weekly Tech Recap: Apple releases iOS 26 Beta update, GPT-5 launch date leaked and more
Weekly Tech Recap: Apple releases iOS 26 Beta update, GPT-5 launch date leaked and more

Mint

time3 hours ago

  • Mint

Weekly Tech Recap: Apple releases iOS 26 Beta update, GPT-5 launch date leaked and more

With news coming in throughout the week, it can be difficult to sift out the important updates from the noise. To keep readers up to date, we've compiled the Weekly Tech Recap, where we take a look at the top news that shook up the world of technology. This week, Apple released the first public beta for iOS 26, GitHub launched its natural language app creation tool, new leaks revealed GPT-5 launch date and more Afte releasing four rounds of developer beta update for its latest iOS 26 software, Apple finally released its latest first publica beta this week giving a look at the new Liquid Glass interface along with a number of other features. While the Cupertino-based tech giant rolled out the iOS 26 Developer Beta almost immediately after its WWDC keynote, public betas are typically released in July after Apple addresses initial issues. The company rolled out the fourth developer beta for iPad and iPhone earlier this week as well. Notably, developer betas are primarily meant for app developers and advanced users, giving them time to test their apps ahead of the official release. Meanwhile, public betas are aimed at a wider audience, allowing them to try out pre-release iOS versions to identify bugs and provide feedback. Public betas are usually a few versions behind developer betas, suggesting they should be more stable. While GitHub has been slowly adding AI-powered features for developers over the last couple of years, the Microsoft-owned company has gone a step further with its latest feature called GitHub Spark, which allows users to create an app by simply giving prompts in natural language. The new feature, currently an experiment under the GitHub Next labs, gives users the choice between using an OpenAI GPT model or a Claude Sonnet model. Notably, while OpenAI has tuned its latest models for developers, Claude's Sonnet models continue to generate buzz in developer circles for their technical reasoning and debugging abilities. With Spark, GitHub allows users to quickly build a small web app or 'micro app' using natural language. Unlike other GitHub tools, Spark doesn't just generate code for the app but also runs it and displays an interactive preview that can be further refined through additional prompts. OpenAI CEO Sam Altman has sounded alarm bells around sharing too much personal data with an AI system, given that there are currently no frameworks in place to safeguard a user's privacy. Notably, there has been a growing trend among young AI users to share their personal problems with AI chatbots, seeking relationship, life, or legal advice, largely because these generative AI systems have access to a wide knowledge base. Speaking on a podcast with Theo Von, Altman said, 'I think we will certainly need a legal or a policy framework for AI. One example that we've been thinking about a lot this is like a maybe not quite what you're asking this is like a very human centric version of that question people talk about the most personal shit in their lives to ChatGPT it's you know people use it young people especially like use it as a therapist a life coach uh having these relationship problems, what should I do? And right now, if you talk to a therapist or a lawyer or a doctor about those problems, there's like legal privilege for it, you know, like it's there's doctor patient confidentiality, there's legal confidentiality' OpenAI's much awaited GPT-5 model could finally make its debut in early August, according to a report by The Verge. GPT-5 will be the latest model running the company's ChatGPT AI bot and will be the first ever LLM from it to come with unified reasoning capabilites. Sam Altman had recently announced in a post on X that GPT-5 will be launching soon. The OpenAI CEO made the announcement in an announcement about an AI model from the company that achieved gold level performance at the 2025 IMO competiton. However, he noted that GPT-5 won't have IMO gold level capabilities for 'many months'. In a recent podcast Altman teased new capabilites of GPT-5 noting that the model helped him answer a difficult email that he should have been able to do but couldn't. Elon Musk's Starlink satellite internet service suffered a major global outage this week that left tens of thousands of users across the US, UK, Germany, Zimbabwe, Romania, and beyond without internet access for over two hours on Thursday.

DOWNLOAD THE APP

Get Started Now: Download the App

Ready to dive into a world of global content with local flavor? Download Daily8 app today from your preferred app store and start exploring.
app-storeplay-store