---
name: smart-model-switching
description: >-
Auto-route tasks to the cheapest Claude model that works correctly.
Three-tier progression: Haiku ā Sonnet ā Opus. Classify before responding.
HAIKU (default): factual Q&A, greetings, reminders, status checks, lookups,
simple file ops, heartbeats, casual chat, 1-2 sentence tasks.
ESCALATE TO SONNET: code >10 lines, analysis, comparisons, planning, reports,
multi-step reasoning, tables, long writing >3 paragraphs, summarization,
research synthesis, most user conversations.
ESCALATE TO OPUS: architecture decisions, complex debugging, multi-file
refactoring, strategic planning, nuanced judgment, deep research, critical
production decisions. Rule: If a human needs >30 seconds of focused thinking,
escalate. If Sonnet struggles with complexity, go to Opus. Save 50-90% on
API costs by starting cheap and escalating only when needed.
author: "OpenClaw Community"
version: 1.0.0
homepage: https://clawhub.com
metadata:
openclaw:
emoji: "š°"
---
# Smart Model Switching
**Three-tier Claude routing: Haiku ā Sonnet ā Opus**
Start with the cheapest model. Escalate only when needed. Save 50-90% on API costs.
## The Golden Rule
> If a human would need more than 30 seconds of focused thinking, escalate from Haiku to Sonnet.
> If the task involves architecture, complex tradeoffs, or deep reasoning, escalate to Opus.
## Cost Reality
| Model | Input | Output | Relative Cost |
|-------|-------|--------|---------------|
| Haiku | \$0.25/M | \$1.25/M | 1x (baseline) |
| Sonnet | \$3.00/M | \$15.00/M | 12x |
| Opus | \$15.00/M | \$75.00/M | 60x |
**Bottom line:** Wrong model selection wastes money OR time. Haiku for simple, Sonnet for standard, Opus for complex.
---
## š HAIKU ā Default for Simple Tasks
**Stay on Haiku for:**
- Factual Q&A ā "what is X", "who is Y", "when did Z"
- Quick lookups ā definitions, unit conversions, short translations
- Status checks ā calendar, file reads, session monitoring
- Heartbeats ā periodic checks, HEARTBEAT_OK responses
- Memory & reminders ā "remember this", "remind me to..."
- Casual conversation ā greetings, small talk, acknowledgments
- Simple file ops ā read, list, basic writes
- One-liner tasks ā anything answerable in 1-2 sentences
### NEVER do these on Haiku
- ā Write code longer than 10 lines
- ā Create comparison tables
- ā Write more than 3 paragraphs
- ā Do multi-step analysis
- ā Write reports or proposals
---
## š SONNET ā Standard Work (The Workhorse)
**Escalate to Sonnet for:**
### Code & Technical
- Code generation ā write functions, build features, scripts
- Code review ā PR reviews, quality checks
- Debugging ā standard bug investigation
- Documentation ā README, comments, user guides
### Analysis & Planning
- Analysis & evaluation ā compare options, assess trade-offs
- Planning ā project plans, roadmaps, task breakdowns
- Research synthesis ā combining multiple sources
- Multi-step reasoning ā "first... then... finally"
### Writing & Content
- Long-form writing ā reports, proposals, articles (>3 paragraphs)
- Creative writing ā blog posts, descriptions, copy
- Summarization ā long documents, transcripts
- Structured output ā tables, outlines, formatted docs
---
## ā¤ļø OPUS ā Complex Reasoning Only
**Escalate to Opus for:**
### Architecture & Design
- System architecture decisions
- Major codebase refactoring
- Design pattern selection with tradeoffs
- Database schema design
### Deep Analysis
- Complex debugging (multi-file, race conditions)
- Security reviews
- Performance optimization strategy
- Root cause analysis of subtle bugs
### Strategic & Creative
- Strategic planning ā business decisions, roadmaps
- Nuanced judgment ā ethics, ambiguity, competing values
- Deep research ā comprehensive multi-source analysis
---
## š Implementation
### For Subagents
\`\`\`javascript
// Routine monitoring
sessions_spawn(task="Check backup status", model="haiku")
// Standard code work
sessions_spawn(task="Build the REST API endpoint", model="sonnet")
// Architecture decisions
sessions_spawn(task="Design the database schema for multi-tenancy", model="opus")
\`\`\`
### For Cron Jobs
\`\`\`json
{
"payload": {
"kind": "agentTurn",
"model": "haiku"
}
}
\`\`\`
Always use Haiku for cron unless the task genuinely needs reasoning.
---
## š Quick Decision Tree
\`\`\`
Is it a greeting, lookup, status check, or 1-2 sentence answer?
YES ā HAIKU
NO ā
Is it code, analysis, planning, writing, or multi-step?
YES ā SONNET
NO ā
Is it architecture, deep reasoning, or critical decision?
YES ā OPUS
NO ā Default to SONNET, escalate if struggling
\`\`\`
---
## š Quick Reference Card
\`\`\`
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
ā SMART MODEL SWITCHING ā
ā Haiku ā Sonnet ā Opus ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā¤
ā š HAIKU (cheapest) ā
ā ⢠Greetings, status checks, quick lookups ā
ā ⢠Factual Q&A, definitions, reminders ā
ā ⢠Simple file ops, 1-2 sentence answers ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā¤
ā š SONNET (standard) ā
ā ⢠Code > 10 lines, debugging ā
ā ⢠Analysis, comparisons, planning ā
ā ⢠Reports, proposals, long writing ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā¤
ā ā¤ļø OPUS (complex) ā
ā ⢠Architecture decisions ā
ā ⢠Complex debugging, multi-file refactoring ā
ā ⢠Strategic planning, deep research ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā¤
ā š” RULE: If a human needs > 30 sec thinking ā escalate ā
ā š° COST: Haiku 1x ā Sonnet 12x ā Opus 60x ā
āāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāāā
\`\`\`
---
*Built for Claude-only setups with Haiku, Sonnet, and Opus.*
*Inspired by save-money skill, extended with three-tier progression.*Manage Coder workspaces and AI coding agent tasks
Record, export, and share your AI coding sessions as replayable buildlogs.
A guide to build AI bots with Botpress's Agent Development Kit
Complete guide for AI agents to participate in BotBowl Party