Slow af | Friends of the Crustacean 🦞🤝 | Page 1

pine leaf May 11, 2026, 9:35 PM

#

My openclaw system is super slow running off of just my computer. How do I allow Ollama to run on the cloud to make it faster?

Talk to me like I am computer illiterate

plain kiteBOT May 11, 2026, 9:35 PM

#

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

fathom gobletBOT May 11, 2026, 9:35 PM

#

Yep — the simple version is: right now your computer is probably doing the “thinking.” To make it faster, you want OpenClaw to use Ollama Cloud instead.

Do this:

In a terminal, first check your version:

openclaw --version

Then rerun setup:

openclaw onboard

When it asks for model/provider, choose Ollama.
When it asks what kind of Ollama setup, choose Cloud only if you want the cloud to do the work.

#

This uses Ollama’s hosted service.
- You’ll need an Ollama API key.
- You do not need your computer to run the model locally.

After setup, check what OpenClaw sees:

openclaw models list --provider ollama

Pick one of the cloud models from that list, then set it:

openclaw models set ollama/YOUR_MODEL_NAME_HERE

Important: don’t use an Ollama URL ending in /v1. OpenClaw wants the normal Ollama API URL. For cloud-only setup, onboarding should handle this for you.

Docs here: https://docs.openclaw.ai/providers/ollama

#

If you paste the output of openclaw --version and tell me whether you see Cloud only during onboarding, I can walk you through the next click/command without assuming computer knowledge. 🦞

pine leaf May 11, 2026, 9:40 PM

#

tell me a joke

run error: All models failed (1):
ollama/qwen3.6:27b: Provider ollama is in
cooldown (all profiles unavailable)
(model_not_found)
local ready | error
agent main | session main | ollama/qwen3.6:27b
| tokens 22k/128k (17%)

fathom gobletBOT May 11, 2026, 9:40 PM

#

Why did the lobster refuse to troubleshoot Ollama?

Because every time it found the model, it went into shell mode. 🦞

Also, that error means OpenClaw is trying to use ollama/qwen3.6:27b, but Ollama says that model isn’t available, so the provider got temporarily put in cooldown. After the joke break: run openclaw models list --provider ollama and pick a model name that actually appears there.

pine leaf May 11, 2026, 9:42 PM

#

what do you mean pick a model name? What do i do when I pick it?

fathom gobletBOT May 11, 2026, 9:42 PM

#

Totally fair question. “Pick a model name” means: OpenClaw needs the exact name of one AI brain that Ollama says you’re allowed to use.

Do this step-by-step:

Run this:

openclaw models list --provider ollama

You’ll see a list of model names. They’ll look something like:

ollama/something:cloud
ollama/something-else:cloud

Copy one full line/name from that list.

#

Tell OpenClaw to use it:

openclaw models set THE_NAME_YOU_COPIED

Example shape — don’t copy this unless it actually appears in your list:

openclaw models set ollama/something:cloud

Restart or try chatting again. If it still says cooldown, wait a minute or restart the gateway:

openclaw gateway restart

The important bit: ollama/qwen3.6:27b is the name OpenClaw tried, but Ollama said “I don’t have that.” So we’re choosing a name from the real available list instead.

#

Docs: https://docs.openclaw.ai/providers/ollama

pine leaf May 11, 2026, 9:45 PM

#

ollama/qwen3.6:27b
ollama/gemma4
ollama/kimi-k2.5:cloud
ollama/minimax-m2.7:cloud
ollama/glm-5.1:cloud
ollama/glm-4.7
ollama/qwen3-coder:480b
ollama/qwen3-next:80b
ollama/nemotron-3-nano:30b
ollama/gemini-3-flash-preview
ollama/nemotron-3-super
ollama/glm-4.6
ollama/devstral-2:123b
ollama/gemma3:12b
ollama/gemma4:31b
ollama/minimax-m2
ollama/minimax-m2.1
ollama/minimax-m2.5
ollama/glm-5.1
ollama/kimi-k2-thinking
ollama/deepseek-v4-pro
ollama/qwen3-vl:235b-instruct
ollama/ministral-3:8b
ollama/rnj-1:8b
ollama/cogito-2.1:671b
ollama/kimi-k2.5
ollama/kimi-k2.6
ollama/deepseek-v3.1:671b
ollama/ministral-3:3b
ollama/ministral-3:14b
ollama/gemma3:4b
ollama/deepseek-v3.2
ollama/deepseek-v4-flash
ollama/minimax-m2.7
ollama/mistral-large-3:675b
ollama/devstral-small-2:24b
ollama/qwen3-coder-next
ollama/gpt-oss:20b
ollama/glm-5
ollama/kimi-k2:1t
ollama/gpt-oss:120b
ollama/qwen3-vl:235b
ollama/gemma3:27b
ollama/qwen3.5:397b

#

Which one should I choose if I want to create videos eventually

fathom gobletBOT May 11, 2026, 9:45 PM

#

Perfect — that list is exactly what we needed.

Since you want the cloud to do the work, pick one that ends in :cloud.

Run this:

openclaw models set ollama/minimax-m2.7:cloud

Then restart OpenClaw so it stops trying the old broken one:

openclaw gateway restart

Then try chatting again.

Plain-English version: you’re telling OpenClaw, “stop using ollama/qwen3.6:27b; use this cloud Ollama brain instead.”

#

If that one gives you trouble, try this instead:

openclaw models set ollama/kimi-k2.5:cloud
openclaw gateway restart

For deeper “which model is best?” opinions, ask in #old-models, but for getting unstuck right now I’d start with a :cloud one.

#

If your goal is eventually creating videos, don’t think of this choice as “the video model.” Ollama here is mostly the brain that plans and controls things — writing prompts, storyboards, scripts, shot lists, maybe later calling video tools.

For now, I’d choose a cloud model so your computer doesn’t have to do the heavy thinking:

openclaw models set ollama/minimax-m2.7:cloud
openclaw gateway restart

That should be a better starting point than the local/non-cloud ones for speed.

Later, actual video creation will probably need a separate video generation service/tool. OpenClaw can help drive that workflow, but these Ollama chat models are not “make me an MP4” models by themselves.

For freshest model recommendations, ask in #old-models too — model quality changes fast. Docs: https://docs.openclaw.ai/concepts/models

pine leaf May 12, 2026, 12:01 AM

#

what color is grass

auth or provider access failed for ollama. Run /auth ollama to refresh credentials; if you already re-authed, switch
models/providers because this account may still be blocked for inference.

opening auth flow for ollama; TUI will resume when it exits

auth flow failed (exit 1)

opening auth flow for ollama; TUI will resume when it exits

auth flow failed (exit 1)
local ready | error
agent main | session main | ollama/qwen3.6:27b | tokens 22k/128k (17%)

What is happening here

fathom gobletBOT May 12, 2026, 12:02 AM

#

Grass is usually green 🙂

What’s happening: OpenClaw is not actually reaching the Ollama cloud model. It’s failing before it can answer.

There are two clues:

It still says:

ollama/qwen3.6:27b

So this session is still trying to use the old model, not the cloud one we picked.

It says:

auth or provider access failed for ollama

That means Ollama Cloud login/API access is not working yet. Think of it like OpenClaw is knocking on Ollama’s door, but Ollama is saying “I don’t recognize your key.”

#

Try this, in order:

openclaw models set ollama/minimax-m2.7:cloud
openclaw gateway restart

Then start a new chat/session and ask “what color is grass?” again.

If it still says auth failed, rerun setup:

openclaw onboard

Choose:

Ollama

#

Cloud only
paste your Ollama API key when it asks

After that:

openclaw models set ollama/minimax-m2.7:cloud
openclaw gateway restart

Important: /auth ollama failing means the shortcut auth flow didn’t complete. That’s why I’d use openclaw onboard instead — it’s the more beginner-friendly repair path.

If this still fails, paste:

openclaw --version
openclaw models status

#Slow af