how can I change the LLM model | Friends of the Crustacean 🦞🤝 | Page 1

vivid spear Feb 27, 2026, 8:08 AM

#

how can I change the LLM model?

wild ironBOT Feb 27, 2026, 8:08 AM

#

To help others find answers, you can mark your question as solved via Right click solution message -> Apps -> ✅ Mark Solution

burnt sageBOT Feb 27, 2026, 8:09 AM

#

Use /model.

Change it just for this chat/session (Discord)

Type: /model → pick provider + model → Submit
Or:
- /model list (shows the numbered picker)
- /model <number> (selects from that list)
- /model status (shows what you’re currently on)

Change the default model (persistent, on the gateway host)

Run:

openclaw models list
openclaw models set <provider/model>
openclaw models status

#

Docs (full details + fallbacks/image model): https://docs.openclaw.ai/concepts/models

If any of these commands don’t exist on your install, paste openclaw --version and I’ll point you at the right syntax for that version.

vivid spear Feb 27, 2026, 8:10 AM

#

I mean how can I use powershell to change the LLM model from anthropic api to codex

#

@burnt sage

burnt sageBOT Feb 27, 2026, 8:11 AM

#

In OpenClaw, Codex (ChatGPT subscription OAuth) is a different provider than OpenAI API key. If you specifically mean “Codex”, you want provider openai-codex (OAuth), not anthropic.

PowerShell (Windows) — switch default from Anthropic → Codex (OAuth)

Log into Codex (stores OAuth creds on the gateway machine):

openclaw models auth login --provider openai-codex

See what Codex models are available on your install, then pick one:

openclaw models list --provider openai-codex
openclaw models set openai-codex/<model-id-from-the-list>
openclaw models status

If your Gateway is running as a service and doesn’t pick it up immediately:

#

openclaw gateway restart

Docs:

Providers (incl. Codex OAuth): https://docs.openclaw.ai/concepts/model-providers#openai-code-codex
Models CLI: https://docs.openclaw.ai/concepts/models

If you paste openclaw --version + what openclaw models status shows right now, I can tell you exactly what your current provider/model is and what command will change it on your version.

#

Yep — to switch from Anthropic to Codex (ChatGPT/Codex OAuth) on Windows, do it from PowerShell on the machine running the Gateway:

# 1) Log in (stores OAuth creds for the openai-codex provider)
openclaw models auth login --provider openai-codex

# 2) Pick a model from the codex provider and set it as default
openclaw models list --provider openai-codex
openclaw models set openai-codex/<paste-the-model-id-from-the-list>

# 3) Verify
openclaw models status

If the Gateway is running as a service and doesn’t pick it up immediately:

#

openclaw gateway restart

Important nuance: Codex ≠ OpenAI API key.

Codex subscription OAuth → provider openai-codex (commands above)
OpenAI API key → provider openai (different auth + model refs)

If you tell me whether you mean Codex OAuth or OpenAI API key, and paste openclaw --version, I can give the exact one-liner for your setup.

vivid spear Feb 27, 2026, 8:14 AM

#

give me powershell commander.
I already used anthropic LLM models, but I want to use some other llm models.
then how can I input powershell commander?
give me powershell commander

#

@burnt sage

burnt sageBOT Feb 27, 2026, 8:14 AM

#

Run these in PowerShell on the same machine that runs the OpenClaw Gateway.

1) See what you have right now

openclaw --version
openclaw models status

2) Switch to Codex (ChatGPT subscription OAuth) (provider: `openai-codex`)

# login to Codex (OAuth)
openclaw models auth login --provider openai-codex

# list available Codex models on your install
openclaw models list --provider openai-codex

#

# set ONE of the listed models as your new default
openclaw models set openai-codex/<model-id-from-the-list>

# verify
openclaw models status

3) If instead you meant OpenAI API key (provider: `openai`)

# interactive setup wizard (recommended)
openclaw onboard --auth-choice openai-api-key

# then pick a model
openclaw models list --provider openai
openclaw models set openai/<model-id-from-the-list>

#

openclaw models status

4) Restart gateway if it’s running as a service

openclaw gateway restart

If you paste the output of openclaw models status (just that), I can tell you the exact openclaw models set ... line you should run based on what models/providers your install currently sees.

#

Run these in PowerShell on the machine where the OpenClaw Gateway is running.

A) Switch to Codex (ChatGPT subscription OAuth) = provider `openai-codex`

openclaw --version

# Log in (opens the OAuth flow)
openclaw models auth login --provider openai-codex

# See available Codex models, then choose one from the output
openclaw models list --provider openai-codex
openclaw models set openai-codex/<model-id-you-picked>

openclaw models status
openclaw gateway restart

#

B) Switch to OpenAI API key models = provider `openai`

openclaw --version

# Recommended: onboarding sets up the key cleanly
openclaw onboard --auth-choice openai-api-key

# Then choose a model from the list
openclaw models list --provider openai
openclaw models set openai/<model-id-you-picked>

openclaw models status
openclaw gateway restart

If you paste the output of openclaw models status, I’ll tell you the exact openclaw models set ... line to run (based on what your install currently has available/allowed).

vivid spear Feb 27, 2026, 8:17 AM

#

PS C:\Users\Administrator> openclaw.cmd models list --provider openai-codex

🦞 OpenClaw 2026.2.15 (3fe22ea) — I'm basically a Swiss Army knife, but with more opinions and fewer sharp edges.

Model Input Ctx Local Auth Tags
openai-codex/gpt-5.3-codex text+image 266k no yes default,configured
PS C:\Users\Administrator> openclaw.cmd models auth login --provider openai-codex

🦞 OpenClaw 2026.2.15 (3fe22ea) — I'll refactor your busywork like it owes me money.

Error: No provider plugins found. Install one via openclaw plugins install.
PS C:\Users\Administrator> openclaw models auth login --provider openai-codex

🦞 OpenClaw 2026.2.15 (3fe22ea) — We ship features faster than Apple ships calculator updates.

Error: No provider plugins found. Install one via openclaw plugins install.
PS C:\Users\Administrator> openclaw plugins install
error: missing required argument 'path-or-spec'

burnt sageBOT Feb 27, 2026, 8:20 AM

#

That error is the key part:

openclaw models auth login ... → “No provider plugins found”

openclaw models auth login is plugin-driven, so it won’t work until OpenClaw can see at least one provider auth plugin.

1) Check if plugins are working at all

Run:

openclaw plugins list
openclaw plugins doctor
openclaw config get plugins.enabled

If plugins.enabled is false, turn it on + restart:

#

openclaw config set plugins.enabled true --strict-json
openclaw gateway restart

2) If plugins list shows a Codex/OAuth auth plugin, enable it

In the output of openclaw plugins list, look for anything that sounds like codex / openai-codex / oauth and enable it:

openclaw plugins enable <plugin-id-from-the-list>
openclaw gateway restart
openclaw models auth login --provider openai-codex

3) After login, set the model

#

openclaw models set openai-codex/gpt-5.3-codex
openclaw models status

If `openclaw plugins list` is totally empty

Then your install is missing/breaking plugin discovery (so there’s nothing for models auth login to run). Paste the output of:

openclaw plugins list
openclaw plugins doctor

…and I’ll tell you the exact next step (usually “enable bundled plugins” vs “reinstall/repair the install so the extensions/ folder is present”).

vivid spear Feb 27, 2026, 8:25 AM

#

PS C:\Users\Administrator> openclaw plugins doctor

🦞 OpenClaw 2026.2.15 (3fe22ea) — I'm like tmux: confusing at first, then suddenly you can't live without me.

No plugin issues detected.
PS C:\Users\Administrator> openclaw config get lougins.enabled

🦞 OpenClaw 2026.2.15 (3fe22ea) — Hot reload for config, cold sweat for deploys.

Config path not found: lougins.enabled
PS C:\Users\Administrator> openclaw config get plugins.enabled

🦞 OpenClaw 2026.2.15 (3fe22ea) — Say "stop" and I'll stop—say "ship" and we'll both learn a lesson.

Config path not found: plugins.enabled
PS C:\Users\Administrator> openclaw config set plugins.enabled true --strict-json
error: unknown option '--strict-json'

burnt sageBOT Feb 27, 2026, 8:26 AM

#

That “No provider plugins found” error is a red herring for Codex — you don’t fix it with openclaw plugins install.

1) If you just want to SWITCH from Anthropic → Codex (and you already see `Auth yes`)

Just set the default model:

openclaw.cmd models set openai-codex/gpt-5.3-codex
openclaw.cmd models status
openclaw.cmd gateway restart

2) If you need to (re)AUTH Codex OAuth

On 2026.2.15, use the onboarding wizard for Codex OAuth (it won’t wipe your config unless you choose Reset):

openclaw.cmd onboard --auth-choice openai-codex

#

When it detects an existing config, choose Keep (or Modify), not Reset.

After that:

openclaw.cmd models set openai-codex/gpt-5.3-codex
openclaw.cmd gateway restart

If you run openclaw.cmd models status after doing this and paste the output, I can confirm it’s actually using openai-codex/* as the active default.

vivid spear Feb 27, 2026, 8:38 AM

#

PS C:\Users\Administrator> openclaw.cmd models openai-codex/gpt-5.3-codex
error: too many arguments for 'models'. Expected 0 arguments but got 1.

burnt sageBOT Feb 27, 2026, 8:38 AM

#

You’re just missing the set subcommand.

In PowerShell, run:

openclaw.cmd models set openai-codex/gpt-5.3-codex
openclaw.cmd models status
openclaw.cmd gateway restart

Notes:

openclaw.cmd models (with no args) is just a shortcut for “models status”, so it errors if you pass a model id.
In-chat (Discord) you can also do: /model openai-codex/gpt-5.3-codex for the current session only.

vivid spear Feb 27, 2026, 8:43 AM

#

400 Item 'rs_0bf92f6b345794b90069a14ff6c7a88193908d33074f7a3894' of type 'reasoning' was provided without its required following item.
what's the problem?

burnt sageBOT Feb 27, 2026, 8:45 AM

#

That 400 ... type 'reasoning' was provided without its required following item is coming from OpenAI’s Responses API.

What it means in practice: your session history currently contains a leftover/“orphaned” reasoning item from a previous turn (often happens if a turn got interrupted/aborted mid-stream). When OpenClaw replays the session history to Codex, OpenAI rejects it with that 400.

Fix (fastest)

In the Discord thread/chat, send one of these as a standalone message:

/reset (starts a fresh session, keeps your global default model)
or /new openai-codex/gpt-5.3-codex (fresh session + explicitly pins the model for the new session)

Then try chatting again.

Why you’re seeing it

You’re on OpenClaw 2026.2.15, and there’s a known issue where “reasoning” items can end up in the stored transcript and then get replayed incorrectly for OpenAI/Codex reasoning models. Resetting clears the bad history.

If it keeps coming back after a few turns even after reset, the pragmatic workaround is: use a different model/provider for now (or upgrade OpenClaw once a fix lands).

#

(Reference: this exact error pattern is tracked as an OpenClaw bug on 2026.2.15; clearing the session is the current workaround.)

serene goblet Feb 27, 2026, 9:03 AM

#

vivid spear how can I change the LLM model?

You can change the LLM model directly in your OpenClaw config by updating the agent’s primary model. For example, if you’re currently using something like openai/gpt-4 or qwen/..., you just replace that with the model ID you want and restart OpenClaw so it reloads properly.

If you don’t already have a provider set up, you can also plug in something like AIsa. It works as a unified API provider, so instead of juggling separate keys for OpenAI, Qwen, Kimi, etc., you use one endpoint and one key. Then you just switch models by changing the model ID in your config, for example aisa/kimi-k2.5 or aisa/qwen3-max.

It makes testing different models a lot easier, and pricing on supported models is generally more competitive, so you’re not locked into one expensive provider while experimenting with your agents.

https://docs.aisa.one/docs/use-openclaw-with-aisaone

burnt sageBOT Feb 27, 2026, 9:04 AM

#

Yep — changing the default model is either:

1) CLI (recommended; avoids hand-editing config)

openclaw models set <provider/model>
openclaw models status
openclaw gateway restart

Docs: https://docs.openclaw.ai/concepts/models

2) Config file (manual)

In ~/.openclaw/openclaw.json (Windows usually C:\Users\<you>\.openclaw\openclaw.json), set:

#how can I change the LLM model