AI Models
OpenClaw works with the supported model IDs you configure. Use the provider and model IDs that match your BYOK setup.
Available Models
| Profile | Example ID | Best For | Speed | Cost |
|---|---|---|---|---|
| Fast | google/gemini-2.5-flash | Simple queries, fast responses | Fastest | Lowest |
| Balanced | openai/gpt-5-mini | Daily assistant work | Fast | Medium |
| Reasoning | openrouter/auto | Complex reasoning and coding | Variable | Variable |
Default Model
Set the default model for all conversations:
AI_MODEL=openrouter/autoModel Routing
Configure automatic model selection based on task type:
# Use a faster model for simple queries and a stronger model for complex ones
MODEL_ROUTING_ENABLED=true
MODEL_ROUTING_SIMPLE=google/gemini-2.5-flash
MODEL_ROUTING_COMPLEX=openai/gpt-5-mini
MODEL_ROUTING_CODE=openrouter/autoPer-Request Model Selection
Users can request a specific configured model when you expose that option in your deployment.
Cost Comparison
Estimated costs per 1,000 tasks (average complexity):
| Profile | Input Cost | Output Cost | ~1K Tasks |
|---|---|---|---|
| Fast | Low | Low | Lowest |
| Balanced | Medium | Medium | Moderate |
| Reasoning | Variable | Variable | Highest |
Recommendations
Personal Use
- Default: Balanced profile
- Fallback: Fast profile for simple queries
- Estimated: $10-30/month
Shared Use
- Default: Balanced profile
- Enable model routing
- Estimated: $50-150/month
Power Users
- Default: Balanced profile
- Reasoning profile for code review
- Estimated: $30-100/month