OpenClaw works with any LLM, and because it acts across your apps through 100+ skills, the best model is one with dependable tool use at a price that survives constant background use. Claude Sonnet 4.6 and GPT-5.4 are the strongest balanced drivers; for value, agent-focused open models like MiniMax M3, MiniMax M2.1, DeepSeek V4 Pro, and GLM 5.2 keep a local-first, always-on agent affordable.
The most dependable tool-using driver — handles OpenClaw's multi-skill workflows cleanly without the flagship price.
Purpose-built for agents: reliable tool use and a 1M context at low cost — a natural fit for OpenClaw's always-on style.
A strong, well-priced generalist when you want OpenAI-grade reasoning behind your skills.
Near-frontier quality at open-weight prices — and self-hostable, which suits OpenClaw's local-first design.
Fast and cheap with a 1M context — great for the high volume of small actions OpenClaw fires off.
A capable open-weight flagship for the harder, multi-step automations.
The model picks the moves; the agent runs the loop, the tools, and the guardrails. Once you've chosen a model, see which agent gets the most out of it.