How to Switch OpenClaw Models in Chat

You do not need to restart the Gateway just to try a different model. OpenClaw has a chat-level model picker, which is exactly what you want when a session needs a stronger model for one task or a cheaper one for routine work. The key is understanding that the switch is session-scoped and still respects the configured model allowlist.

When this is the right move

Use chat switching when you are inside a conversation and the next task changes shape: a long reasoning problem, a coding pass, a quick summarization, or a multimodal task. Use config changes instead when you want to change the default for future sessions or all agents. If the selected model is blocked, fix agents.defaults.models rather than repeatedly retrying the same chat command.

The practical workflow

1. Type /model or /model list to open the compact picker. On Discord, the docs say /model and /models open an interactive picker with provider and model dropdowns plus submit.
2. Pick by number when the model appears in the list, or type the full provider/model reference when you know it exactly.
3. Use /model status when you need detail about auth candidates or configured provider endpoint information.
4. If a run is idle, the next turn uses the new model right away. If a run is already active, OpenClaw queues the switch until a clean retry opportunity or later turn.
5. If you get a not-allowed message, add the model to the allowlist, clear the allowlist, or choose from the allowed picker options.

Grounded command or config pattern

These are the documented chat commands for model switching.

/model
/model list
/model 3
/model openai/gpt-5.4
/model status

Model refs split on the first slash, so provider/model is the shape to preserve. The docs also note that /model persists the new session selection immediately, which means you are changing this conversation, not just the next single reply.

Operator notes

This is useful operationally because it separates default policy from moment-to-moment judgment. I like strong defaults for tool-enabled sessions, then explicit switches for low-risk summarization or high-context analysis. The history stays in the same session, but the runtime model can move when the work demands it.

Rollout approach

For switch openclaw models in chat, I would make the first pass deliberately small: one owner, one machine or channel, one visible test, and one rollback path. OpenClaw features become powerful when they connect to real tools and real messages, so the safest rollout is not a giant configuration day. It is a short rehearsal that proves the docs-grounded path works in your exact workspace before you depend on it while busy.

Common mistake

The common mistake is treating the command as the whole feature. The command starts the workflow, but the surrounding state is what keeps it reliable: config validation, auth, pairing, permissions, logs, and a tiny verification step. If those pieces are skipped, the next failure looks random even when OpenClaw is behaving exactly as configured.

Maintenance rhythm

Once this is working, write down the exact command, config path, or approval decision you used. Future you will not remember the tiny detail that made the setup safe. A small note in the workspace or runbook is cheaper than rediscovering the same behavior during an outage, especially after updates or machine changes.

Safety checks

Do not use model switching as a way around policy. If agents.defaults.models is intentionally constrained, respect that. For untrusted inputs, tool-heavy sessions, or anything that can touch files, messages, browser state, or shell commands, prefer the strongest trusted model available rather than chasing pennies on a fragile model.

How to verify it worked

After switching, run /model status or ask a small status-style question. If the next reply still uses the old model, there may have been an active run and the switch was pending, or the session selected a model that auth cannot actually use.

If you want the operator version with sharper checklists, safer defaults, and fewer “why is this broken?” afternoons, The OpenClaw Playbook is the shortcut I would hand to a serious OpenClaw owner.