Switching Your Agent's AI Model
Your agent comes pre-configured with a great default model, but you can switch anytime to balance cost, speed, and quality.
Why Switch Models?
Save Credits
Budget models cost 10–20× less. For routine tasks, a cheaper model works just as well.
Get More Speed
Flash-tier models respond faster, great for quick back-and-forth conversations.
Upgrade Quality
For complex writing or reasoning, premium models like Claude Sonnet deliver noticeably better results.
How to Switch
Just tell your agent
The simplest way — just say it in chat:
You: switch to openrouter/anthropic/claude-sonnet-4.5
→ Agent switches model and confirms
Use the /model command
Send a slash command with the model ID:
/model openrouter/minimax/minimax-m2.5
/model openrouter/moonshotai/kimi-k2.5
/model openrouter/anthropic/claude-sonnet-4.5
Browse available models
See all models your agent can use:
/models
→ Lists all available models with current pricing
Check current model
Not sure which model you're on?
/status
→ Shows current model and session info
Model Naming Format
OpenClaw uses the openrouter/<author>/<slug> format for all OpenRouter models. For example:
# Anthropic models
openrouter/anthropic/claude-sonnet-4.5
openrouter/anthropic/claude-haiku-4.5
# OpenAI models
openrouter/openai/gpt-4.1-mini
openrouter/openai/gpt-4.1-nano
# Other providers
openrouter/moonshotai/kimi-k2.5
openrouter/z-ai/glm-5
openrouter/minimax/minimax-m2.5
# Auto router — picks the best model for each prompt
openrouter/openrouter/auto
The openrouter/openrouter/auto option automatically selects the most cost-effective model for each message. Great if you want to let OpenRouter optimize for you. Browse all 300+ models at openrouter.ai/models.
Available Models
All models are accessed via OpenRouter. Prices are per 1M tokens (input / output).
MiniMax M2.5
MiniMaxDEFAULTYour default model. Excellent quality at the best price — strong coding, reasoning, and conversation.
openrouter/minimax/minimax-m2.5Kimi K2.5
Moonshot AIFALLBACKDefault fallback model. Strong reasoning with the cheapest input pricing available.
openrouter/moonshotai/kimi-k2.5GLM-5
Z-AISolid performance across a wide range of tasks at a great value price point.
openrouter/z-ai/glm-5GPT-4.1 Mini
OpenAIReliable and predictable. A great choice for straightforward tasks and automation.
openrouter/openai/gpt-4.1-miniClaude Sonnet 4.5
AnthropicPREMIUMPremium quality for complex reasoning, nuanced writing, and advanced analysis.
openrouter/anthropic/claude-sonnet-4.5Claude Haiku 4.5
AnthropicFast Anthropic model. Good for quick tasks where you want Claude's quality.
openrouter/anthropic/claude-haiku-4.5Qwen 3.5 Plus
QwenStrong multilingual capabilities. Excellent for non-English languages.
openrouter/qwen/qwen3.5-plusCredit Impact
Approximate messages per $5 of OpenRouter credits (based on ~500 input + ~500 output tokens per message):
| Model | ~Msgs per $5 | Input / Output |
|---|---|---|
| MiniMax M2.5default | ~2,800 | $0.30 / $1.20 |
| GPT-4.1 Mini | ~2,200 | $0.40 / $1.60 |
| Kimi K2.5 | ~1,600 | $0.23 / $3.00 |
| GLM-5 | ~1,500 | $0.30 / $2.55 |
| Qwen 3.5 Plus | ~1,500 | $0.40 / $2.40 |
| Claude Haiku 4.5 | ~720 | $1.00 / $5.00 |
| Claude Sonnet 4.5 | ~240 | $3.00 / $15.00 |
Estimates assume ~500 input tokens + ~500 output tokens per message. Real usage varies based on conversation length and response size. Complex tasks with long context will consume more tokens.
BYOK: Use Any OpenRouter Model
On any plan, you can bring your own OpenRouter API key and access hundreds of additional models — including GPT-4o, Llama 3, Mistral, and more. Credits are billed directly by OpenRouter at their rates.
Pro Tip: Start with the Default
Your agent ships with minimax/minimax-m2.5 as primary and moonshotai/kimi-k2.5 as fallback. This combo is carefully chosen for the best balance of cost and quality. Only switch if you have a specific reason — like needing premium quality for a complex task.
Ready to try a different model?
Open your Telegram chat and type /model openrouter/anthropic/claude-sonnet-4.5 to switch right now.