How to chat with Apple Intelligence on your iPhone: setup, tools and limits

Apple's own place to chat with Apple Intelligence on iOS 27 is the new Siri app that comes with Siri AI, an English beta with a waitlist that is not available on iPhones in the EU. Apple also lets other apps use its on-device model through the Foundation Models framework. This guide covers how to chat with Apple Intelligence in Orvena: what you need, where to switch it on, what the model can do with your calendar, maps and photos, what works offline, and where it runs out of room.

Where you can chat with Apple Intelligence in iOS 27

Apple Intelligence is a set of features rather than an app: Writing Tools, summaries, Live Translation, Genmoji, Image Playground and photo editing, some running on the iPhone and some on Apple's servers. Apple's app for conversations is the new Siri app that comes with Siri AI, and Siri AI is an English beta with a waitlist, and Apple says it "will not be available initially in the EU in iOS, iPadOS, and watchOS."

Apple also opens the on-device model to other apps. Its Apple Intelligence page says: "Any app can tap into the on-device models that power Apple Intelligence, and the features you build work offline. And it's all at no cost per request." The route is the Foundation Models framework, which "works on any Apple Intelligence-compatible device when Apple Intelligence is enabled." iOS 27 ships a newer version of the model that, in Apple's words, "follows instructions more accurately and produces better results, including in complex scenarios." Several App Store apps put a chat window on it. Orvena uses it as one of its engines, with the same phone tools as its own model.

What you need

It works on iPhones in the EU. Only Siri AI is held back there; Apple says the new Apple Intelligence features across apps "are available for users based in the EU with iOS 27", and Orvena 2.0.0 was tested with Apple Intelligence on iPhones in Denmark. What EU iPhone users get and miss is covered separately.

How to use Apple Intelligence in Orvena without a download

During setup, the model page, "Better answers, still private.", shows an Apple Intelligence card when the iPhone runs iOS 27 in a language Apple supports. Depending on the state, its caption reads "Built into iOS. Ready now, with nothing to download.", "Turn it on in the iPhone's Settings to start without a download." or "iOS is still getting it ready on this iPhone." When it is ready, tap Use. If no other model is set up it is already checked, and you can ask your first question with no download.

Later, open Models with the model button at the top right of a conversation, or from the menu at the top left, then Settings, then Models. The card sits under Built in. Tap Use, and a checkmark marks it as active; Orvena's downloaded model, if you have one, stays on the phone but is unloaded from memory. The caption changes with the state:

An eye icon next to the name means the model understands images, so you can attach a photo. The default setting, Automatic, uses Orvena's downloaded model when it is loaded and otherwise Apple Intelligence, if it supports your iPhone's language. Use picks Apple Intelligence either way.

What Apple's on-device model can do with Orvena's tools

Orvena gives Apple's model the same tools as its own:

• Calendar: read your events, add one, move one. Reminders: add, list, complete.
• Alarms and timers.
• Maps: travel time and distance, place details, places nearby, and opening directions in Maps.
• Weather, from Apple's weather service.
• Photos: search saved photos by date, album, favorite, type or place.
• Health: steps, sleep and workouts.
• Files you added to Orvena, contacts, Music and your Shortcuts.
• Web search through DuckDuckGo, and reading a web page.
• With Premium Unlock: scheduled tasks, and connections such as Notion, GitHub or your own MCP server.

An answer from Orvena 2.0.0 on Apple Intelligence: asked "How long is the walk from the Colosseum to the Trevi Fountain in Rome?", it showed a travel-time card, Colosseum to Trevi Fountain, 1.6 km, about 20 min, with the route on a map, and answered "The walk from the Colosseum to the Trevi Fountain in Rome is about 20 minutes, covering 1.6 km." Place lookups in 2.0.0 find landmarks in other cities, so you do not have to be in Rome for it to pick the right Colosseum.

On iOS 27, when Orvena is not on screen and the iPhone is unlocked, "Ask Orvena" in Siri, Shortcuts or Spotlight is answered where you asked by Apple Intelligence, whether Apple Intelligence or Orvena's downloaded model is your engine, as long as it supports your iPhone's language. An answer in progress keeps going if you switch apps, until iOS ends Orvena's background time.

What works offline with Apple Intelligence

The model runs on the iPhone, so answering needs no connection. Orvena's privacy policy says: "When Orvena answers with Apple Intelligence, Apple's on-device model runs on the iPhone as well, and Orvena sends those requests nowhere else." In airplane mode conversation keeps working, and so do calendar, reminders, alarms, timers, photos, Health, contacts and files.

Tools that reach the outside world need a connection: web search goes to DuckDuckGo, a web page to that site, weather and travel times to Apple's services, and connections to the service you signed in to. The Rome route above needed the network; the model did not. Offline, these tools say they cannot reach the service.

Context window: Apple Intelligence, Orvena's model and Private Cloud Compute

The context window is how much of a conversation, tool results included, a model can read at once. Apple's documentation still says "Apple's on-device foundation model has a context window of 4096 tokens per session", but on the iPhones Orvena 2.0.0 was tested on with iOS 27, the model reported 8,192 tokens. The size can differ between iPhones, so Orvena does not hard-code a number. It reads the size from iOS at runtime, as Apple asked developers to do to "adapt your app to the hardware it's running on."

Apple Intelligence languages

Apple's list for iOS 27 has sixteen: "English, Danish, Dutch, French, German, Italian, Norwegian, Portuguese, Spanish, Swedish, Turkish, Vietnamese, Chinese (simplified), Chinese (traditional), Japanese, and Korean." Siri AI has a separate, shorter list: English now, with French, Japanese, Korean, Portuguese and Spanish due in October.

Automatic only picks Apple Intelligence when it supports your iPhone's language, and you can still choose it with Use. If Apple's model rejects the language of a message, Orvena shows "Apple Intelligence doesn't support this language yet. Download Orvena's own model in Models to chat in it." Orvena's own model works "in any language it knows", as the Models page puts it, and Orvena's voice speaks 31 languages, among them Finnish, Polish, Czech, Greek, Ukrainian, Russian, Arabic, Hindi and Indonesian, none of which are on Apple's list.

Limits

• No Thinking setting. The Thinking row on the Models page, Off or Automatic, is hidden while Apple Intelligence is active. It only applies to Orvena's own model.
• Shorter conversations. When a conversation outgrows the window, Orvena keeps the most recent whole turns and leaves out the oldest; it does not summarize them. If even that does not fit, you see "This conversation is too long for Apple Intelligence. Start a new chat, or switch to Orvena's downloaded model in Models."
• Shorter tool results. A tool result is cut to 2,400 characters before Apple's model reads it, against 4,000 with Orvena's own model, so a long web page or file is shortened sooner.
• Apple's guardrails. Apple's framework applies its own safety checks, and the model can also refuse. When either stops a reply, Orvena shows "Apple Intelligence declined to answer this. Try rephrasing it, or switch to Orvena's downloaded model in Models."

When something else is the better choice

Orvena's own model, Qwen 3.5 4B, is a one-time 3.1 GB download under On-device models, and it runs on the phone too. It is the better choice for a language Apple does not support, for long conversations and documents, with 16,384 tokens of context, for control over Thinking, and on iOS 26. Both can stay installed, and you switch with Use.

Private Cloud Compute is Apple's larger model on its servers, part of Premium Unlock on iOS 27, within a daily limit Apple sets and iCloud+ raises. Apple says it has "stronger reasoning for handling long documents or extended multiturn conversations." Asked "Suggest a walking tour of Rome for the day, starting at the Colosseum, with walking times between the stops", it searched the web, read a page, and made a travel-time card for each leg, from the Colosseum past the Arch of Constantine, Trajan's Column and Piazza Venezia to the Capitoline Hill, 24 minutes of walking in all. It needs a connection, and each answer sends the conversation to Apple's servers, so Orvena asks once when you choose it, and every time before health or financial data goes. What Private Cloud Compute is covers Apple's guarantees.

Apple's own assistant does some things Orvena does not. Where Siri AI is available, Apple says it "can understand personal context across messages, emails, photos, and more" and can "answer questions related to the content on a user's screen", and its app syncs conversations across your devices through iCloud.

Common questions

Is there an Apple Intelligence app to chat with on iPhone?

Where Siri AI is available, its new Siri app is where you chat with Apple Intelligence, but Siri AI is an English beta with a waitlist and is not on iPhones in the EU. Other apps can use Apple's on-device model through the Foundation Models framework; in Orvena you pick it on the Models page, under Built in.

Does Apple Intelligence work offline?

The on-device model does. In Orvena it answers and uses the calendar, reminder, photo and Health tools with no connection. Web search, weather, maps and Private Cloud Compute need one.

Which iPhones can run Apple Intelligence?

Apple lists iPhone 15 Pro and 15 Pro Max, every iPhone 16 including the 16e, iPhone 17, 17e, Air, 17 Pro and 17 Pro Max, iPhone 18 Pro and 18 Pro Max, and iPhone Duo. Orvena needs iOS 27 to use it.

Which languages does Apple Intelligence support?

Sixteen on iOS 27: English, Danish, Dutch, French, German, Italian, Norwegian, Portuguese, Spanish, Swedish, Turkish, Vietnamese, Chinese (simplified and traditional), Japanese and Korean. For other languages, Orvena's own model is the one to use.

Does Apple Intelligence work in the EU?

Yes. Siri AI is not available on iPhones in the EU at launch, but Apple says the other new Apple Intelligence features are, and Orvena's Apple Intelligence engine was tested on iPhones in Denmark running iOS 27.