Rate Limited — All Models Failed Cooldown

Fix OpenClaw rate limit and cooldown errors when every model fallback is temporarily unavailable because of provider limits, quota, billing, missing auth, or overload.

Run Openclaw Now

Quick Fix

Run these commands first:

openclaw status
openclaw gateway status
openclaw models status
openclaw logs --follow

Use this only when you need a live provider check:

openclaw models status --probe

Do not run --probe again and again during a rate limit. It can make real provider checks, and repeated checks may add more pressure instead of helping. OpenClaw docs list openclaw models status, auth checks, and model config checks as useful ways to confirm model availability.

What This Error Means

OpenClaw can fail over in two main steps:

It can try another auth profile for the same provider.
It can try the next model in your configured fallback list.

If all available routes fail, OpenClaw shows an “all models failed” or cooldown error.

Common error messages include:

HTTP 429
Rate limit exceeded
Too many requests
RESOURCE_EXHAUSTED
Quota exceeded
All models failed
No available auth profile
Provider is in cooldown
retry-after

The final “all models failed” message is usually only a summary. The first real error in the logs tells you the real cause.

Most Common Causes

1. Provider Rate Limit

This is the most common reason.

AI providers limit usage by requests, tokens, projects, organizations, and sometimes model-specific limits.

OpenAI rate limits can include requests per minute, requests per day, tokens per minute, tokens per day, and other model-specific limits.

Anthropic Claude API rate limits are measured by requests per minute, input tokens per minute, and output tokens per minute. If you exceed a limit, Anthropic returns a 429 error with a retry-after value.

Google Gemini rate limits are applied per project and can vary by model. Google also notes that requests per day reset at midnight Pacific time.

How to check

openclaw logs --follow
openclaw models status
openclaw gateway status --deep

Fix

Wait for the retry or cooldown time.
Reduce repeated requests.
Reduce long prompts.
Stop retry loops.
Check provider billing and quota.
Add a fallback from another provider or account.

2. All Fallbacks Use the Same Provider

Fallbacks will not help much if every fallback uses the same exhausted provider, account, or quota pool.

Weak fallback setup:

{
  "agents": {
    "defaults": {
      "model": {
        "primary": "provider-a/main-model",
        "fallbacks": [
          "provider-a/backup-model-1",
          "provider-a/backup-model-2"
        ]
      }
    }
  }
}

Better fallback setup:

{
  "agents": {
    "defaults": {
      "model": {
        "primary": "provider-a/main-model",
        "fallbacks": [
          "provider-b/available-model",
          "provider-c/available-model"
        ]
      }
    }
  }
}

Use real model names from your own OpenClaw setup. Check them with:

openclaw models status

A good fallback should be:

Already authenticated
Available in your provider account
Not blocked by the same quota pool
Strong enough for the task
Tested before production use

3. Your Session Is Locked to One Model

If you selected a model inside a chat session, OpenClaw may treat that as a strict choice.

That means fallback may not work the way you expect.

Check the selected model:

/model status

Start a fresh session:

/new

Or reset the session:

/reset

OpenClaw docs explain that explicit user session selections can be strict, while configured defaults and fallback policies can use configured fallbacks.

4. Long Chat or Huge Prompt

Long sessions can trigger rate limits faster because OpenClaw may send a lot of previous context.

This can happen with:

Long conversations
Large files
Big logs
Tool output
Images
PDFs
Repeated automation runs

Fix

Start a new session.
Split large tasks into smaller tasks.
Remove unnecessary logs.
Avoid uploading the same large file again.
Use a model that can handle the request size.
Use smaller models for simple background jobs.

5. Too Many Automated Requests

OpenClaw may hit cooldown faster if automations run too often.

Common examples:

Cron job runs every minute
Discord bot responds to every message
Webhook keeps retrying
Multiple channels trigger the same workflow
Several long tasks run at the same time

Fix

Increase cron intervals.
Stop retry loops.
Use mention-only rules in busy group chats.
Use cheaper models for background tasks.
Do not run too many long tasks at once.

6. Billing or Credits Are Exhausted

Some errors look like rate limits but are actually billing problems.

Look for:

insufficient credits
billing limit
monthly spend limit
quota exceeded
organization spending limit exceeded

Fix

Open the provider dashboard.
Check billing.
Add credits or payment method.
Raise spend limits if needed.
Use another provider/account with available quota.

OpenClaw cannot fix a provider account that has no credits or blocked billing. Annoying, yes. Magic, no.

7. Provider Is Temporarily Overloaded

Sometimes the provider is busy.

You may see:

overloaded
server overloaded
upstream error
timeout
model not ready
internal server error

Fix

Wait a few minutes.
Try a smaller or faster model.
Use a fallback provider.
Check provider status.
Do not restart OpenClaw repeatedly unless the gateway itself is unhealthy.

Check gateway health:

openclaw gateway status
openclaw logs --follow

Provider vs OpenClaw Problem

What You See	Likely Cause	What To Do
HTTP 429	Provider rate limit	Wait, reduce usage, check provider dashboard
RESOURCE_EXHAUSTED	Google/Gemini quota	Check Google project quota
No available auth profile	Auth profile missing or cooling down	Run `openclaw models status`
insufficient credits	Billing issue	Add credits or update billing
Gateway not running	OpenClaw gateway issue	Run `openclaw gateway status`
Long chat fails only	Context too large	Start new session
Fallback also fails	Bad fallback setup	Add tested fallback from another provider

Step-by-Step Fix

Step 1: Check OpenClaw Status

openclaw status
openclaw gateway status

If the gateway is not running, fix that first.

Step 2: Read Logs

openclaw logs --follow

Look for the first provider error.

Do not only read the final “all models failed” line. That is usually just the summary.

Step 3: Check Model Status

openclaw models status

Check:

Primary model
Fallback models
Auth profile status
Missing auth
Cooldown state
Selected session model

Step 4: Check Provider Dashboard

Check the dashboard for the provider shown in the error:

OpenAI: usage, project limits, org limits
Anthropic: Claude Console limits and spend limits
Google Gemini: AI Studio limits and project quota

If the provider dashboard shows no available quota, OpenClaw cannot fix that locally.

Step 5: Add a Real Fallback

Use at least one fallback from another provider or account.

Example:

{
  "agents": {
    "defaults": {
      "model": {
        "primary": "provider-a/main-model",
        "fallbacks": [
          "provider-b/available-model",
          "provider-c/available-model"
        ]
      }
    }
  }
}

Then verify:

openclaw models status

What Not To Do

Avoid these mistakes:

Do not keep sending the same prompt every few seconds.
Do not restart the gateway repeatedly if the provider is rate limited.
Do not use fallbacks that all share the same exhausted provider.
Do not ignore billing or credit errors.
Do not run --probe repeatedly during cooldown.
Do not assume “all models failed” means OpenClaw is broken.
Do not run high-frequency automations on expensive long-context models.

Prevention Checklist

Before running heavy OpenClaw workflows:

Run openclaw models status
Add fallback models from another provider/account
Keep provider billing active
Avoid very long sessions
Split large tasks
Reduce cron frequency
Stop retry loops
Use smaller models for background jobs
Monitor provider dashboards
Use managed hosting if uptime matters

Frequently Asked Questions

Why does OpenClaw say all models failed?

It means OpenClaw tried the available model routes, but none could complete the request.

Is this always an OpenClaw bug?

No. It is usually caused by provider rate limits, quota, billing, missing auth, or bad fallback setup.

How long should I wait?

Wait for the retry time or cooldown shown in the error. If billing or quota is exhausted, waiting may not help.

Why did fallback not work?

Fallback may fail if it uses the same provider, same quota pool, missing auth, or unavailable model.

Can restarting OpenClaw fix it?

Restarting can help if the gateway is stuck or config changed. It does not refill provider quota or fix billing.

How do I check the real cause?

Run: openclaw logs --follow, openclaw models status, and openclaw gateway status. Read the first provider error in the logs.

Does Ampere.sh prevent all rate limits?

No. Provider rate limits still apply. Ampere.sh helps with managed OpenClaw hosting, uptime, setup, and operations.

Also Read

Guide

OpenClaw Bot Not Responding? Fix API Key & Gateway Issues

Guide

How to Change Model in OpenClaw: Complete Guide for Beginners

10 min·May 25, 2026

Guide

How to Reduce OpenClaw API Cost Without Losing Workflow Quality

Written by

Emma Thompson

AI Research Writer

Emma is an AI researcher and technical writer with a PhD in Machine Learning from Stanford. She specializes in large language model evaluation, comparing model capabilities, and explaining complex AI concepts. Her research has been published in NeurIPS and ICML. She makes cutting-edge AI research accessible through clear, practical guides.

Stop fighting model cooldowns

Managed OpenClaw hosting helps with uptime, setup, gateway health, and day-to-day operations. Provider rate limits still apply, but the OpenClaw side is easier to manage.

Start 7-Day Free Trial →