Apple Intelligence

Apple Intelligence brings fast, private, on-device AI to your iPad and Mac workflows. Built directly into your device's processor, it provides instant AI capabilities without internet, API costs, or cloud processing.

What is Apple Intelligence?

Apple Intelligence is Apple's on-device AI system that runs entirely on your iPad or Mac. When you select Apple Intelligence as your AI model in Circuitry, your workflows use the powerful neural engine built into your device's M-series chip—no data leaves your device, and there are no usage fees.

Key Benefits:

• 💰 Free - No API costs or subscription fees
• ⚡ Fast - 30+ tokens/second with no network latency
• 🔒 Private - All processing happens on your device
• ✈️ Offline - Works without internet connection
• 🎯 Optimized - Built specifically for Apple Silicon

How to Use Apple Intelligence

In Agent Nodes

1. Add an Agent node to your workflow
2. In the node configuration panel, find the Model dropdown
3. Select "Apple Intelligence" (appears first when available)
4. Configure your prompt as usual
5. Run your workflow

Apple Intelligence works exactly like other AI models—just select it from the dropdown and use it as you would GPT-4 or Claude.

In AI Wizards

Apple Intelligence is available in all AI wizards:

• Code Wizard (Cmd/Ctrl+W in Code nodes) - For data transformations and simple scripts
• Data Wizard (in DataGrid nodes) - Excellent for entity extraction
• Web Wizard (in Web nodes) - For content summarization
• Agent Wizard (Cmd/Ctrl+W in Agent nodes) - For prompt refinement

When available, Apple Intelligence appears as the first option in the model selector within each wizard.

Automatic Detection

Circuitry automatically detects if your device supports Apple Intelligence. If you're using:

• iPad Pro (M1, M2, M4, M5 or later)
• iPad Air (M1, M2 or later)
• Mac (M1 or later)
• iPadOS 26+ or macOS 26+

...then Apple Intelligence will be available and prioritized in all model selectors.

Capabilities: When to Use Apple Intelligence

Apple Intelligence excels at specific types of tasks. Use this guide to choose the right model for your workflow.

Capabilities Matrix

✅ Use Apple Intelligence For:

Summarization

• Condensing long documents, notes, or meeting transcripts
• Creating bullet-point summaries
• Extracting key points from text

Entity Extraction

• Extracting names, dates, emails, phone numbers
• Pulling structured data from unstructured text
• Parsing receipts, invoices, forms

Classification & Tagging

• Categorizing customer feedback
• Auto-tagging content by topic
• Sentiment analysis
• Content moderation

Quick Data Transformations

• Reformatting data
• Simple data cleaning
• Converting between formats

Privacy-Sensitive Workflows

• Medical/health data processing
• Legal document analysis
• Personal information handling
• Any data that shouldn't leave your device

Offline Workflows

• Field work without internet
• Airplane mode
• Areas with poor connectivity

❌ Use Cloud Models (GPT-4, Claude, etc.) For:

Complex Reasoning

• Multi-step logic problems
• Strategic analysis
• Advanced decision-making

World Knowledge

• Questions requiring up-to-date information
• Historical facts
• General knowledge queries

Long-form Content Generation

• Blog posts, articles
• Documentation
• Creative stories
• Marketing copy

Advanced Code Generation

• Complex algorithms
• Full application code
• Architectural decisions

Conversational AI

• Extended chat interactions
• Context-heavy discussions

Real Workflow Examples

Example 1: Email to Contact List

Use Case: Extract contacts from a batch of emails

Best Model: Apple Intelligence ⭐⭐⭐⭐⭐

[Start Node]
  → [Agent Node: Apple Intelligence]
     Prompt: "Extract all contacts with name, email, phone from this text"
  → [DataGrid Node]
     Display: Contact list in spreadsheet

Why Apple Intelligence:

• Entity extraction is a core strength
• Fast processing for large email batches
• No API costs for bulk operations
• Works offline during travel

Example 2: Meeting Notes Summary

Use Case: Summarize meeting notes into action items

Best Model: Apple Intelligence ⭐⭐⭐⭐⭐

[Start Node]
  → [Agent Node: Apple Intelligence]
     Prompt: "Summarize this meeting into 5 bullet points with action items"
  → [Message Node]
     Display: Summary to user

Why Apple Intelligence:

• Summarization is optimal use case
• Instant results without API latency
• Private meeting notes stay on device
• Free for unlimited meeting summaries

Example 3: Customer Feedback Analysis

Use Case: Categorize customer feedback by sentiment and topic

Best Model: Apple Intelligence ⭐⭐⭐⭐⭐

[Start Node]
  → [Loop Node: Iterate over feedback]
     → [Agent Node: Apple Intelligence]
        Prompt: "Classify this feedback as positive/negative/neutral and extract main topic"
     → [DataGrid Node: Collect results]

Why Apple Intelligence:

• Classification is a strength
• Fast batch processing
• No per-item API costs
• Keeps customer data private

Example 4: Research Report Generation

Use Case: Generate comprehensive research report with citations

Best Model: GPT-4 or Claude Sonnet ⭐⭐⭐⭐⭐

[Start Node]
  → [Agent Node: GPT-4]
     Prompt: "Research X and write a 2000-word report with citations"
  → [Web Node: Display formatted report]

Why Cloud Model:

• Requires world knowledge
• Long-form content generation
• Complex reasoning and synthesis
• Apple Intelligence not designed for this

Example 5: Code Generation for API Integration

Use Case: Generate code to integrate with a complex REST API

Best Model: GPT-4 or Claude Sonnet ⭐⭐⭐⭐⭐

[Start Node]
  → [Agent Node: GPT-4]
     Prompt: "Generate code to integrate with the Stripe API for payment processing"
  → [Code Node: Execute or review code]

Why Cloud Model:

• Complex code generation
• Requires API-specific knowledge
• Multi-file code structures
• Apple Intelligence better for simple scripts

Example 6: Hybrid Approach - Expense Report Processing

Use Case: Extract expenses from receipts, then generate summary report

Best Model: Both! ⭐⭐⭐⭐⭐

[Start Node]
  → [Agent Node: Apple Intelligence]
     Prompt: "Extract vendor, amount, date, category from this receipt"
  → [DataGrid Node: Collect all expenses]
  → [Agent Node: GPT-4]
     Prompt: "Analyze spending patterns and write executive summary"
  → [Web Node: Display report]

Why Hybrid:

• Apple Intelligence: Fast, free extraction (strength)
• Cloud Model: Complex analysis and reporting (strength)
• Cost-effective: Only pay for final summary, not per-receipt extraction
• Best of both worlds

Cost Savings

Apple Intelligence can dramatically reduce your AI costs for supported tasks:

Example Monthly Costs:

For workflows that process large volumes of data requiring summarization, extraction, or classification, Apple Intelligence can save hundreds of dollars per month while providing faster results.

Privacy & Security

On-Device Processing

When you use Apple Intelligence in Circuitry:

• All AI processing happens on your device - your data never leaves your iPad or Mac
• No cloud uploads - unlike cloud AI models that send data to servers
• No data retention - Apple doesn't store or log your requests
• No external API calls - completely self-contained

Ideal for Sensitive Data

Apple Intelligence is perfect for workflows involving:

• Healthcare data (HIPAA compliance)
• Legal documents (attorney-client privilege)
• Financial records (PCI/SOX compliance)
• Personal information (GDPR compliance)
• Trade secrets and proprietary information
• Any data that can't or shouldn't be sent to cloud services

Regulatory Compliance

Organizations with strict data handling requirements can use Apple Intelligence for AI-powered workflows while maintaining compliance, as data never leaves the device.

Offline Workflows

When Offline Capability Matters

Apple Intelligence works completely offline, making it ideal for:

Field Work

• Medical professionals visiting patients
• Sales representatives on the road
• Field service technicians
• Real estate agents at properties

Travel

• Airplane mode during flights
• International travel without roaming
• Remote locations with no connectivity
• Poor network coverage areas

Data Center Operations

• Air-gapped environments
• Secure facilities without internet
• Network-isolated systems

Reliability

• Always available, regardless of internet status
• No API rate limits or outages
• No dependency on third-party services

Example: Field Service Workflow

Scenario: Technician inspects equipment and generates report offline

[Start Node]
  → [Form Node: Collect inspection data]
  → [Agent Node: Apple Intelligence]
     Prompt: "Analyze inspection data and identify issues"
  → [Agent Node: Apple Intelligence]
     Prompt: "Generate service report with recommendations"
  → [Message Node: Display report]
  → [Action: Save to local database]

Later, when online:
  → [Action: Sync to cloud]

All AI processing happens on the iPad, no internet required.

Device Requirements

Compatible Devices

iPad:

• iPad Pro 11-inch (2021 or later) - M1, M2, M4, M5
• iPad Pro 12.9-inch (2021 or later) - M1, M2, M4, M5
• iPad Pro 13-inch (2024 or later) - M4, M5
• iPad Air 11-inch (2024 or later) - M2
• iPad Air 13-inch (2024 or later) - M2

Mac:

• Any Mac with M1, M2, M3, M4, or M5 chip
• iMac (2021 or later)
• Mac mini (2020 or later)
• MacBook Air (2020 or later)
• MacBook Pro (2020 or later)
• Mac Studio (2022 or later)
• Mac Pro (2023 or later)

System Requirements

• iPadOS 26 or later (for iPad)
• macOS 26 or later (for Mac)
• Apple Intelligence enabled in Settings
• 7GB of available storage for on-device model
• 8GB of RAM minimum (most M-series devices have this)

Checking Availability

Circuitry automatically detects if Apple Intelligence is available on your device. If you don't see "Apple Intelligence" in your model selector:

1. Check your device is on the compatible list above
2. Update to iPadOS 26+ or macOS 26+
3. Go to Settings → Apple Intelligence and turn it on
4. Ensure you have 7GB of free storage
5. Restart Circuitry

Regional Availability

Apple Intelligence is currently available in:

• United States
• United Kingdom
• Australia
• Canada
• Ireland
• New Zealand
• South Africa

Supported languages:

• English (US, UK, Australia, Canada, Ireland, New Zealand, South Africa)
• French (France)
• German (Germany)
• Italian (Italy)
• Portuguese (Brazil)
• Spanish (Spain, Mexico, US)
• Chinese (Simplified)
• Japanese
• Korean

Performance Expectations

Speed

Apple Intelligence:

• Token generation: ~30 tokens/second
• Latency: Near-instant (0.6ms per input token)
• No network delay: Processes immediately on device

Cloud Models (for comparison):

• Token generation: 10-50 tokens/second (varies by model and server load)
• Latency: 500ms-2s network round-trip before processing begins
• Network dependent: Slower on poor connections

Response Quality

Excellent:

• Summarizing up to ~2,000 words
• Extracting 10-100+ entities
• Classifying into 2-50 categories
• Structured data output (JSON, lists, tables)

Good:

• Short creative writing (1-2 paragraphs)
• Simple data transformations
• Basic code snippets (10-50 lines)

Limited:

• Long-form content (500+ words)
• Complex multi-step reasoning
• World knowledge questions
• Advanced code generation

Troubleshooting

"Apple Intelligence" Not Showing in Model List

Possible Causes:

1. Device not compatible - Check Device Requirements section
2. System version too old - Update to iPadOS/macOS 26+
3. Apple Intelligence not enabled - Go to Settings → Apple Intelligence
4. Insufficient storage - Free up at least 7GB
5. Region not supported - Check Regional Availability section

Solution:

• Circuitry will automatically show Apple Intelligence when all requirements are met
• Use cloud models (GPT-4, Claude) as alternatives

Apple Intelligence Unavailable During Workflow Execution

If Apple Intelligence becomes unavailable mid-workflow:

Automatic Fallback:

• Circuitry will display an error message
• Re-run the workflow, or
• Edit the Agent node to select a different model

Common Causes:

• Apple Intelligence disabled in Settings
• System update in progress
• Device storage critically low

Slow Performance

If Apple Intelligence seems slower than expected:

Check:

1. Device temperature - Thermal throttling can slow processing
2. Available RAM - Close other apps if device is under memory pressure
3. Storage space - Ensure 10GB+ free (not just 7GB minimum)
4. Background processes - Heavy background tasks can impact speed

Expected Speed:

• Simple summarization: 1-2 seconds
• Entity extraction (100 items): 2-4 seconds
• Classification (batch): 3-5 seconds

If significantly slower, restart your device.

Quality Issues

If Apple Intelligence responses are not meeting expectations:

For Better Results:

1. Use for intended tasks - Check Capabilities Matrix
2. Simplify prompts - Be specific and concise
3. Break into steps - Use multiple Agent nodes for complex workflows
4. Consider cloud models - For tasks outside Apple Intelligence strengths

Example - Poor:

Prompt: "Research the history of automation and write a comprehensive essay
analyzing its impact on society, including economic, social, and technological
factors, with citations from academic sources."

→ Use GPT-4 or Claude instead (requires world knowledge + long-form)

Example - Good:

Prompt: "Summarize these customer reviews into 5 main themes with
example quotes for each theme."

→ Perfect for Apple Intelligence (summarization + extraction)

Tips & Best Practices

1. Match Tasks to Strengths

Review the Capabilities Matrix and use Apple Intelligence for what it does best:

• ✅ Summarize before reasoning - Use Apple Intelligence to summarize data, then GPT-4 to analyze
• ✅ Extract then generate - Use Apple Intelligence to extract entities, then Claude to write content
• ✅ Classify then act - Use Apple Intelligence to categorize, then appropriate model for next step

2. Batch Processing for Cost Savings

For high-volume extraction/classification:

[Start Node]
  → [Loop Node: 10,000 items]
     → [Agent Node: Apple Intelligence] ← Free, fast processing
     → [DataGrid: Collect results]

This workflow costs $0 with Apple Intelligence vs $100+ with cloud models.

3. Hybrid Workflows

Combine models based on their strengths:

[Apple Intelligence: Extract data] → [GPT-4: Complex analysis] → [Apple Intelligence: Summarize results]

Use free Apple Intelligence for preprocessing and postprocessing, expensive cloud models only where needed.

4. Optimize for Offline

Design workflows to work offline by:

• Using Apple Intelligence for all AI tasks
• Caching data locally
• Syncing results when connection available

5. Privacy-First Design

For sensitive workflows:

• Use Apple Intelligence exclusively (no cloud)
• Avoid hybrid approaches that might leak data
• Clearly mark workflows as "On-Device Only"

6. Prompt Engineering

Apple Intelligence responds well to:

• Clear, specific instructions - "Extract name, email, phone"
• Structured output requests - "Return as JSON with these fields"
• Examples - "Like this: {example}"
• Concise prompts - Avoid long preambles

Apple Intelligence responds poorly to:

• Vague requests - "Tell me about this"
• Multiple complex steps - Break into separate nodes
• Requests requiring world knowledge - Switch to cloud model

Frequently Asked Questions

Does Apple Intelligence cost anything to use?

No. Apple Intelligence is completely free—no API fees, no subscription, no usage limits. It's included with your device.

How is Apple Intelligence different from ChatGPT or Claude?

Apple Intelligence is optimized for specific tasks (summarization, extraction, classification) and runs entirely on your device. ChatGPT and Claude are larger models with broader capabilities but require internet and API fees.

Think of Apple Intelligence as a specialized tool for fast, private data processing, while ChatGPT/Claude are general-purpose AI assistants.

Can I use Apple Intelligence and cloud models in the same workflow?

Yes! Many workflows benefit from combining models. Use Apple Intelligence for extraction/summarization (fast and free), then cloud models for complex reasoning or generation.

Does using Apple Intelligence drain my battery?

Apple Intelligence is optimized for Apple Silicon and uses minimal power. For typical tasks (summarization, extraction), battery impact is negligible. Long-running batch processing will use more power, similar to any intensive task.

What happens if I start a workflow with Apple Intelligence and lose internet?

The workflow continues running! Apple Intelligence works completely offline. However, if your workflow later needs to access a website or API, that step will fail without internet (same as any workflow).

Can I use Apple Intelligence on my iPhone?

Currently, Circuitry supports Apple Intelligence on iPad and Mac only. iPhone support may be added in the future.

Why doesn't Apple Intelligence appear on my M1 Mac?

Check that you:

1. Have macOS 26 or later installed
2. Have Apple Intelligence enabled in System Settings
3. Have at least 7GB of free storage
4. Are in a supported region
5. Have restarted Circuitry after enabling Apple Intelligence

Is Apple Intelligence HIPAA compliant?

Apple Intelligence processes all data on-device, which helps with HIPAA compliance. However, HIPAA compliance involves many factors beyond the AI model. Consult your compliance team about using any AI in healthcare workflows.

Can Apple Intelligence access the internet?

No. Apple Intelligence runs entirely on your device and has no internet access. It cannot look up current information, fetch web pages, or access online resources. For tasks requiring internet data, use cloud models or add HTTP nodes to your workflow.

What languages does Apple Intelligence support?

English, French, German, Italian, Portuguese, Spanish, Chinese (Simplified), Japanese, and Korean. Language availability depends on your region.

How do I maximize cost savings with Apple Intelligence?

1. Use Apple Intelligence for all summarization and extraction tasks
2. Use batch processing for large datasets
3. Reserve cloud models only for complex reasoning or generation
4. Design hybrid workflows (Apple Intelligence preprocessing → cloud model main task → Apple Intelligence postprocessing)

Summary

Apple Intelligence brings powerful on-device AI to iPad and Mac users, enabling:

• 💰 Zero-cost AI for summarization, extraction, and classification
• ⚡ Instant processing with no network delays
• 🔒 Complete privacy with on-device processing
• ✈️ Offline capability for field work and travel
• 🎯 Optimized performance on Apple Silicon

Use Apple Intelligence when you need: Fast, private, offline AI for summarization, data extraction, classification, or structured output.

Use cloud models when you need: Complex reasoning, world knowledge, long-form content generation, or advanced code generation.

Best practice: Combine both—use Apple Intelligence for what it does best (saving costs), and cloud models where their advanced capabilities are needed.

Start using Apple Intelligence in your workflows today by selecting it from any Agent node or AI wizard model dropdown!