Your AI. Your way.

Writing, understanding a topic or researching an idea can call for different approaches. Nebius Chat lets you choose a model that fits and keep your conversations together in one iOS / iPadOS / macOS app.

At WWDC26, Apple opened its Foundation Models framework to third-party LLMs, running locally or in the cloud. That gives developers a new way to bring several models into one experience. I created Nebius for Foundation Models to connect Nebius open models, then built Nebius Chat to turn that possibility into a product you can use.

This independent project brings together the framework and the application, from the initial idea to the interface. The aim: make choosing an AI feel as natural as choosing a tool for the task.

A clearer explanation, an unexpected idea or a tone that feels right: two models can take different approaches. Give them the same question and context, then compare their answers.

Choose the answer that helps you most. It becomes the starting point for your next message, with no copying between apps. Apple’s local model can also join the comparison on compatible devices.

Nebius Chat also supports iPhone Duo. The outer display suits quick exchanges; opening the device gives you more room to read and compare answers. The interface adapts to the available space.

The same attention carries through to iPhone, iPad and Mac: a native iOS / iPadOS / macOS app with conversations synchronized through your private iCloud storage. Your conversations stay with you, in an interface made for each device.

Plan a weekend, discover a topic or follow an idea further. Web research brings information into your conversation. You can see the searches and pages being read, then find the sources beneath the answer.

See an interesting link? Open the original document and make up your own mind. Powered by Tavily, web research is optional: turn it on when you need it, with a compatible model.

A vocabulary card, an action list or information to reuse in a product: an answer does not always need to be a paragraph. The lab lets you try specific formats and inspect the result field by field.

For developers, it is a hands-on introduction to structured generation with Foundation Models. Describe the format in Swift, send your request and get a reusable result. The framework handles the connection to Nebius.

Which models do you use most? What do your conversations cost? A dashboard brings together your activity, response times and estimated costs by day, week or month, so you can make informed choices.

Amounts are estimates, not invoices, and exclude Tavily research fees. Figures shown in this case study are demonstration data.

Use your own Nebius connection or sponsored access when available. You can also choose Apple’s local model on a compatible device. The app keeps conversations on your devices and in your private iCloud storage, with connection keys stored separately in Keychain.

The choice is explicit: a remote model or web search sends the necessary content to the relevant service. Sponsored access also retains technical records for security and quotas. The privacy policy explains this processing.

Nebius for Foundation Models gives developers a bridge to Nebius models. Nebius Chat shows what that enables in everyday use: compare, research, structure and continue a conversation on iOS / iPadOS / macOS, including iPhone Duo.

My work connected these possibilities into one coherent experience: choosing the use cases, designing the interface and building the framework behind it. An independent initiative, carried from idea to product.

The project shown uses the iOS 27, iPadOS 27 and macOS 27 SDKs. Features depend on the chosen model, connection and device. Localized screenshots show demonstration conversations.