비공개 클라우드 컴퓨팅에서 [Model Name] 활용하기

비공개 클라우드 컴퓨팅을 사용하면 사용자 개인정보를 보호하면서 강력한 최첨단 모델에 접근할 수 있습니다. 이러한 모델의 작동 방식과 Foundation Models 프레임워크를 사용하여 해당 모델에 접근하는 방법을 살펴보세요. 앱에서 사용 가능 여부를 확인하고 원활한 대안을 처리하는 모범 사례를 알아보세요.

챕터

0:00 - Introduction
1:23 - What is Private Cloud Compute
2:43 - Integrating PCC with Foundation Models
4:00 - Deciding between on-device and PCC
4:32 - Reasoning levels and context size
6:15 - Evaluating and combining models
7:10 - Handling usage limits
10:15 - Next steps

리소스

2:49 - Prompt the on-device model

import FoundationModels

  let session = LanguageModelSession()
  let response = try await session.respond(to: "Summarize this article: \(article)")

3:02 - Switch to the PCC server model (one-line change)

import FoundationModels
  
  let session = LanguageModelSession(
      model: PrivateCloudComputeLanguageModel()
  )
  let response = try await session.respond(to: "Summarize this article: \(article)")

3:25 - Structured output and tools work the same

import FoundationModels

  @Generable
  struct ArticleSummary {
      let oneLineSummary: String
      let keyPoints: [String]
  }

  struct FindRelatedArticlesTool: Tool {

  }
  
  let session = LanguageModelSession(
      model: PrivateCloudComputeLanguageModel(),
      tools: [FindRelatedArticlesTool.self]
  )

  let response = try await session.respond(
      to: "Summarize this article: \(article)",
      generating: ArticleSummary.self
  )

3:51 - Check availability

import FoundationModels
  
  struct ArticleSummarizationView: View {
      private var model = PrivateCloudComputeLanguageModel()

      var body: some View {
          if model.isAvailable {
              // Show UI for making request
          } else {
              // Fall back
          }
      }
  }

5:26 - Set a reasoning level

let response = try await session.respond(
      to: prompt,
      contextOptions: ContextOptions(reasoningLevel: .light)
  )
  // Reasoning levels: .light, .moderate, .deep

5:58 - Read the context size

SystemLanguageModel().contextSize
  // 4096 on 26.0
  // 8192 on 27.0 (newer devices)

  PrivateCloudComputeLanguageModel().contextSize
  // 32768

9:41 - Handle usage limits

struct ArticleSummarizationView: View {
      private var model = PrivateCloudComputeLanguageModel()

      var body: some View {
          if case .belowLimit(let info) = model.quotaUsage.status {
              if info.isApproachingLimit {
                  Text("Nearing usage limit.")
                      .foregroundStyle(Color.orange)
              }
          }
          if model.quotaUsage.isLimitReached {
              Text("Usage limit exceeded.")
                  .foregroundStyle(Color.red)
          }
          if let suggestion = model.quotaUsage.limitIncreaseSuggestion {
              Button("Show options") {
                  suggestion.show()
              }
          }
      }
  }

0:00 - Introduction
Access to a new server LLM via Private Cloud Compute. The on-device model also improves this year (image input, better instruction following and tool calling), but PCC enables more complex features: reasoning over large input, many tool calls with large outputs, even from watchOS.
1:23 - What is Private Cloud Compute
PCC delivers a powerful server model without compromising privacy: data is never stored, used only for the request, and independently verified. It's integrated with the OS and iCloud, so there's no authentication or API keys, no token cost to developers, a daily per-user limit (higher with iCloud+), and eligibility for apps under 2M downloads.
2:43 - Integrating PCC with Foundation Models
Prompting the on-device model takes three lines; switching to the PCC server model changes just one. The unified Swift API means Generable structured output and tool calling work identically, so you can switch models without rewriting code, and should check the availability API for non-Apple Intelligence devices.
4:00 - Deciding between on-device and PCC
Both offer privacy, but the on-device model works offline with no request limits and a 4K context, while PCC needs a connection, has a daily limit, offers a 32K context, and supports reasoning.
4:32 - Reasoning levels and context size
Reasoning lets the model think before responding by generating extra transcript text, at three levels (light, moderate, deep). Set it on respond, observe the transcript to show progress, and remember reasoning consumes tokens against the context limit, now readable via the contextSize property.
6:15 - Evaluating and combining models
Choose models and reasoning levels based on data, not vibes; the updated on-device model may surprise you. Use the new Evaluations framework (see "Meet the Evaluations framework") and combine on-device and server models together (see "Build agentic app experiences with Foundation Models").
7:10 - Handling usage limits
Handle the per-user iCloud quota gracefully: check isLimitReached on the model's quotaUsage and show persistent, actionable UI (such as a disabled button with an upgrade option) rather than an alert. Detect the approaching-limit case too, and use Xcode's Simulate Apple Foundation Models Availability debug option to test both states.
10:15 - Next steps
Apply for the server model on the developer website, and explore related content: "What's new in the Foundation Models framework" for an overview and "Debug and profile agentic app experiences with Instruments" for runtime behavior.

시작하기 탐색

알림 받기

플랫폼 탐색

피처링

기술 탐색

피처링

커뮤니티 탐색

피처링

문서 탐색

릴리즈 노트

다운로드 탐색

피처링

지원 탐색

피처링

빠른 링크

비공개 클라우드 컴퓨팅에서 [Model Name] 활용하기

챕터

리소스