Models
What is active, what is available, and when to use each model.
Current model
GPT-5.4
Main conversation and highest-quality general work
OpenAI CodexCloudPaid
This is the current configured default model in OpenClaw.
GPT-5.4
Best current general-use model
OpenAI CodexCloudPaid
Use for serious work, planning, writing, and high-value tasks.
MiniMax M2.5
Cheaper background and bulk work
OpenRouterCloudPaid
Useful when cost matters more than having the strongest model.
Llama 3.3 70B
Local and private background tasks
Ollama / LocalLocalFree
Good for cost control, offline-style workflows, and experimentation.
Gemini
Optional future provider
GoogleCloudUnknown
Not currently configured in this setup. Would require separate provider/API setup.
Recommended routing
How to think about model use
Use GPT-5.4 when
- the work matters
- you want the strongest answer
- it involves planning, writing, or important decisions
Use cheaper or local models when
- the task is repetitive or bulk
- cost matters more than peak quality
- you want private or local-first workflows