Foundation Models On Device

FoundationModels: On-Device LLM (iOS 26)

Patterns for integrating Apple's on-device language model into apps using the FoundationModels framework. Covers text generation, structured output with @Generable, custom tool calling, and snapshot streaming — all running on-device for privacy and offline support.

When to Activate

• Building AI-powered features using Apple Intelligence on-device
• Generating or summarizing text without cloud dependency
• Extracting structured data from natural language input
• Implementing custom tool calling for domain-specific AI actions
• Streaming structured responses for real-time UI updates
• Need privacy-preserving AI (no data leaves the device)

Core Pattern — Availability Check

Always check model availability before creating a session:

struct GenerativeView: View {
    private var model = SystemLanguageModel.default

    var body: some View {
        switch model.availability {
        case .available:
            ContentView()
        case .unavailable(.deviceNotEligible):
            Text("Device not eligible for Apple Intelligence")
        case .unavailable(.appleIntelligenceNotEnabled):
            Text("Please enable Apple Intelligence in Settings")
        case .unavailable(.modelNotReady):
            Text("Model is downloading or not ready")
        case .unavailable(let other):
            Text("Model unavailable: \(String(describing: other))")
        }
    }
}

Core Pattern — Basic Session

// Single-turn: create a new session each time
let session = LanguageModelSession()
let response = try await session.respond(to: "What's a good month to visit Paris?")
print(response.content)

// Multi-turn: reuse session for conversation context
let session = LanguageModelSession(instructions: """
    You are a cooking assistant.
    Provide recipe suggestions based on ingredients.
    Keep suggestions brief and practical.
    """)

let first = try await session.respond(to: "I have chicken and rice")
let followUp = try await session.respond(to: "What about a vegetarian option?")

Key points for instructions:

• Define the model's role ("You are a mentor")
• Specify what to do ("Help extract calendar events")
• Set style preferences ("Respond as briefly as possible")
• Add safety measures ("Respond with 'I can't help with that' for dangerous requests")

Core Pattern — Guided Generation with @Generable

Generate structured Swift types instead of raw strings:

1. Define a Generable Type

@Generable(description: "Basic profile information about a cat")
struct CatProfile {
    var name: String

    @Guide(description: "The age of the cat", .range(0...20))
    var age: Int

    @Guide(description: "A one sentence profile about the cat's personality")
    var profile: String
}

2. Request Structured Output

let response = try await session.respond(
    to: "Generate a cute rescue cat",
    generating: CatProfile.self
)

// Access structured fields directly
print("Name: \(response.content.name)")
print("Age: \(response.content.age)")
print("Profile: \(response.content.profile)")

Supported @Guide Constraints

• .range(0...20) — numeric range
• .count(3) — array element count
• description: — semantic guidance for generation

Core Pattern — Tool Calling

Let the model invoke custom code for domain-specific tasks.

The Tool protocol has two associated types: Arguments (a @Generable type the model fills in) and Output (anything conforming to PromptRepresentable — String, a @Generable type, or GeneratedContent). call(arguments:) returns Output directly. There is no ToolOutput type in the shipping SDK — it existed only in early iOS 26 seeds; Transcript.ToolOutput is an unrelated transcript-entry struct.

1. Define a Tool

struct RecipeSearchTool: Tool {
    let name = "recipe_search"
    let description = "Search for recipes matching a given term and return a list of results."

    @Generable
    struct Arguments {
        var searchTerm: String
        var numberOfResults: Int
    }

    func call(arguments: Arguments) async throws -> String {
        let recipes = await searchRecipes(
            term: arguments.searchTerm,
            limit: arguments.numberOfResults
        )
        return recipes.map { "- \($0.name): \($0.description)" }.joined(separator: "\n")
    }
}

2. Create Session with Tools

let session = LanguageModelSession(tools: [RecipeSearchTool()])
let response = try await session.respond(to: "Find me some pasta recipes")

3. Handle Tool Errors

do {
    let answer = try await session.respond(to: "Find a recipe for tomato soup.")
} catch let error as LanguageModelSession.ToolCallError {
    print(error.tool.name)
    if case .databaseIsEmpty = error.underlyingError as? RecipeSearchToolError {
        // Handle specific tool error
    }
}

Core Pattern — Snapshot Streaming

Stream structured responses for real-time UI with PartiallyGenerated types. The stream yields ResponseStream<Content>.Snapshot elements — the partial value lives in snapshot.content (with snapshot.rawContent holding the raw GeneratedContent):

@Generable
struct TripIdeas {
    @Guide(description: "Ideas for upcoming trips")
    var ideas: [String]
}

let stream = session.streamResponse(
    to: "What are some exciting trip ideas?",
    generating: TripIdeas.self
)

for try await snapshot in stream {
    // snapshot.content: TripIdeas.PartiallyGenerated (all properties Optional)
    print(snapshot.content)
}

SwiftUI Integration

@State private var partialResult: TripIdeas.PartiallyGenerated?
@State private var errorMessage: String?

var body: some View {
    List {
        ForEach(partialResult?.ideas ?? [], id: \.self) { idea in
            Text(idea)
        }
    }
    .overlay {
        if let errorMessage { Text(errorMessage).foregroundStyle(.red) }
    }
    .task {
        do {
            let stream = session.streamResponse(to: prompt, generating: TripIdeas.self)
            for try await snapshot in stream {
                partialResult = snapshot.content
            }
        } catch {
            errorMessage = error.localizedDescription
        }
    }
}

Key Design Decisions

Best Practices

• Always check model.availability before creating a session — handle all unavailability cases
• Use instructions to guide model behavior — they take priority over prompts
• Check isResponding before sending a new request — sessions handle one request at a time
• Access response.content for results — not .output
• Break large inputs into chunks — the context window covers instructions + prompt + output combined; read it from model.contextSize (4,096 today, back-deployed to iOS 26.0) instead of hardcoding, and on iOS 26.4+ measure inputs with model.tokenCount(for:) (overloads for prompt, instructions, tools, schema, and transcript entries)
• Use @Generable for structured output — stronger guarantees than parsing raw strings
• Use GenerationOptions(temperature:) to tune creativity (higher = more creative)
• Monitor with Instruments — use Xcode Instruments to profile request performance
• Re-test prompts after OS updates — the on-device model itself changes with OS releases: Apple documents three model versions so far (26.0–26.3, 26.4, 27.0) and explicitly tells you to re-verify behavior on each ("Updating prompts for new model versions"). The 27 SDK also reorganizes errors — LanguageModelError, SystemLanguageModel.Error, LanguageModelSession.Error (none exist in the 26.x SDK) — and relaxes guardrails that over-blocked benign content

Anti-Patterns to Avoid

• Creating sessions without checking model.availability first
• Sending inputs exceeding the model's context window — or hardcoding 4,096 instead of querying contextSize
• Attempting concurrent requests on a single session
• Using .output instead of .content to access response data
• Parsing raw string responses when @Generable structured output would work
• Building complex multi-step logic in a single prompt — break into multiple focused prompts
• Assuming the model is always available — device eligibility and settings vary
• Assuming prompts tuned on one OS model version behave the same on the next — they may not; re-test on each model version

When to Use

• On-device text generation for privacy-sensitive apps
• Structured data extraction from user input (forms, natural language commands)
• AI-assisted features that must work offline
• Streaming UI that progressively shows generated content
• Domain-specific AI actions via tool calling (search, compute, lookup)