Intelligent Model Selection

Intelligent Model Selection is SpecWeave's automatic system for choosing the right AI model (Claude Haiku, Sonnet, or Opus) based on task complexity. This optimizes both cost and quality by using cheaper models for simple tasks and more capable models for complex ones.

How It Works

Model Tiers

Model	Cost	Best For	Use Cases
Haiku	$	Quick tasks	File search, simple questions, log parsing
Sonnet	$$	Standard work	Planning, code review, implementation
Opus	$$$	Complex tasks	Architecture, multi-file refactoring, novel problems

Phase Detection

SpecWeave detects the current phase to select the appropriate model:

Phase	Typical Model	Why
Exploration	Haiku	Fast file search, codebase navigation
Research	Haiku/Sonnet	Reading docs, understanding context
Planning	Sonnet	Generating specs, breaking down work
Architecture	Sonnet/Opus	Design decisions, ADRs
Implementation	Sonnet	Writing code, tests
Review	Sonnet	Code review, validation
Debugging	Sonnet/Opus	Complex problem solving

Cost Impact

Intelligent model selection can reduce AI costs by 40-60%:

Without intelligent selection:
  10 tasks × Opus = 10 × $15 = $150

With intelligent selection:
  3 research tasks × Haiku = 3 × $0.25 = $0.75
  5 standard tasks × Sonnet = 5 × $3 = $15
  2 complex tasks × Opus = 2 × $15 = $30
  Total: $45.75 (70% savings)

Configuration

Model selection is automatic but can be influenced:

// .specweave/config.json
{
  "ai": {
    "default_model": "sonnet",
    "use_intelligent_selection": true,
    "prefer_cost_optimization": true
  }
}

Override Model Selection

Force a specific model when needed:

# Force Opus for complex task
/specweave:do --model opus

# Use Haiku for simple search
Task agent with model: haiku

How It Works​

Model Tiers​

Phase Detection​

Cost Impact​

Configuration​

Override Model Selection​

Related Terms​