[FEATURE] Automatic model suggestions based on query complexity

Preflight Checklist

• [x] I have searched existing requests and this feature hasn't been requested yet
• [x] This is a single feature request (not multiple features)

Problem Statement

Problem Statement

Currently, users must manually decide when to switch between Opus, Sonnet, and Haiku models. This creates several issues:

1. New users don't know when to switch - They waste expensive Opus credits on simple tasks or struggle with Haiku on complex problems
2. Cognitive burden - Experienced users must constantly evaluate "is this task Opus-worthy?"
3. Suboptimal results - Users often realize they needed a different model only after getting incomplete results
4. Quota waste - Using the wrong model wastes both time and usage limits

Proposed Solution

Proposed Solution

Automatic model suggestions that analyze the user's query and proactively recommend model switches before execution.

Example Flow:

User types: "Design database schema for multi-entity fee management with audit trails"

Claude Code shows:
┌─────────────────────────────────────────────────────────┐
│ 💡 SUGGESTION: Switch to Opus 4.6                      │
│                                                         │
│ Why: Complex architecture + compliance requirements    │
│ Cost: ~5-10 requests on your plan                     │
│                                                         │
│ [Switch & Continue] [Stay with Sonnet]                │
└─────────────────────────────────────────────────────────┘

After architecture is done:

┌─────────────────────────────────────────────────────────┐
│ ✓ Design complete!                                     │
│ 💡 Switch to Sonnet for implementation?                │
│                                                         │
│ [Switch to Sonnet] [Stay with Opus]                   │
└─────────────────────────────────────────────────────────┘

For simple tasks:

User: "Fix typo in variable name"

┌─────────────────────────────────────────────────────────┐
│ ⚡ TIP: Haiku can handle this 3x faster                │
│ [Switch to Haiku] [No thanks]                         │
└─────────────────────────────────────────────────────────┘

Key Features

1. Smart Detection

Analyze queries for:

• Complexity indicators: "architecture", "design", "security", "complex"
• File count (>3 files suggests higher complexity)
• Previous model struggles
• Keywords: financial, legal, compliance, multi-system
• Simple patterns: "typo", "format", "rename", "update copyright"

2. Suggestion Levels

• 🔴 Strong: "This requires Opus - Sonnet may struggle"
• 🟡 Moderate: "Consider Sonnet for better results"
• 🟢 Optional: "Haiku could handle this faster/cheaper"

3. Learning System

• Track user acceptance/rejection patterns
• Adapt suggestions to individual workflow
• Reduce frequency as user builds intuition
• Auto-switch for power users who consistently accept

4. Non-intrusive UX

• Show suggestion before execution (preventive)
• 3-second timeout, then fade
• Keyboard shortcuts: y/n
• Dismissible with Esc
• Option: /suggestions off for power users

Use Case: School ERP Development

I'm building a self-hosted school ERP managing 419+ students with fee structures from ₹16K-37.9K annually. Here's how suggestions would help:

Session workflow:

1. "Add English Lab fee component" → Sonnet (no suggestion needed)
2. "Design security for delete operations" → 💡 Suggest Opus (financial + security)
3. "Implement the PIN form" → 💡 Suggest Sonnet (execution phase)
4. "Fix button alignment" → ⚡ Suggest Haiku (simple edit)

Over time, Claude Code learns I prefer Opus for security/financial features and Sonnet for implementation.

Benefits

For Users:

• Better results with optimal model selection
• Cost savings through efficient quota usage
• Educational - builds model selection intuition
• Reduced frustration from wrong model choice

For Anthropic:

• Reduced support tickets about quota usage
• Increased user satisfaction and retention
• Competitive differentiation (no other AI tool does this)
• Data insights on actual model requirements vs benchmarks
• Upsell opportunities to higher tiers

Implementation Suggestion

Phase 1 (Quick Win):

• Basic keyword detection
• Ship and learn from real usage

Phase 2 (ML-based):

• Train on usage patterns
• Predict model fit from query + context

Phase 3 (Personalization):

• Learn individual user patterns
• Adaptive/automatic switching for power users

Related Issues

This complements the existing opusplan feature (#9749) but extends the concept to user-initiated suggestions rather than just plan/execute mode.

Priority

Medium-High - This addresses a fundamental UX gap in model selection that affects all users, especially those managing quota limits.

---

Environment:

• User Type: School administrator building custom ERP
• Current Usage: Mix of Opus (architecture), Sonnet (development), Haiku (simple edits)
• Pain Point: Constantly deciding which model to use mid-workflow

Alternative Solutions

_No response_

Priority

Critical - Blocking my work

Feature Category

CLI commands and flags

Use Case Example

_No response_

Additional Context

_No response_