429 ComfyUI Error | Rate Limit Causes And Fixes

You’ll usually see a 429 comfyui error when ComfyUI reaches out to something outside your PC and that service says “too many requests.” It can happen during model or image downloads, while pulling previews, or when a custom node hits an API on every queue run.

The good news is that 429 is rarely “broken forever.” It’s a traffic jam. Your job is to stop piling cars onto the on-ramp, then set up a calmer flow so it stays clear for good.

What A 429 Error Means In ComfyUI

HTTP status 429 is the web’s way of telling a client to slow down. In ComfyUI, the “client” might be a browser panel, a manager plugin, a downloader node, or a script that fires multiple requests at once.

Many services add a wait hint in the response headers. If you can see a Retry-After value in logs or dev tools, treat it like a timer. If you don’t see one, assume the limit window is short and step back for a minute.

Places Where 429 Shows Up

• Model downloads — A manager fetches many files fast and triggers throttling on the host.
• Preview images — A gallery view loads lots of thumbnails in one burst.
• API-backed nodes — A node calls an endpoint per prompt, per step, or per batch.
• Shared IP setups — Multiple machines behind one router hit the same limit together.

How To Identify The Throttling Service

If you don’t know which server is sending 429, you’ll keep guessing. The host name is usually visible in the same line as the error, or in the URL of the failed request.

• Check the terminal log — ComfyUI prints request URLs for many nodes and managers.
• Open browser dev tools — In the Network tab, sort by Status and click the 429 row.
• Look at plugin logs — Managers often keep their own log window with the source domain.

Once you have the domain, you can search that provider’s rate limit rules. Don’t guess at limits that you can read in one minute.

Quick Checks Before You Change Anything

Start with the simple stuff. It saves time, and it keeps you from “fixing” the wrong layer.

1. Read the full message — Look for the host name in the log so you know who is rate-limiting you.
2. Retry once after a pause — Close the noisy tab, wait 60–120 seconds, then try again.
3. Check for parallel downloads — If several items are fetching at the same time, cancel the extras.
4. Restart the browser tab — A stuck UI can spam retries even when you stop clicking.
5. Verify your network path — VPNs and shared proxies can push you into stricter limits.

If the error vanishes after a short pause, you’re dealing with a burst limit. If it keeps coming back at the same step, it’s usually a repeated request pattern in the workflow.

Common Triggers That Cause 429 In ComfyUI Workflows

Most 429 problems come from a small set of patterns. Once you spot the pattern, the fix gets straightforward.

Too Many Requests From One Button Click

Some nodes or managers spin up multiple requests per click. If the service counts each file, thumbnail, or metadata call, you can hit the limit in seconds.

• Lower concurrency — Set download threads to 1–2 when pulling big packs.
• Disable auto-preview — Turn off thumbnail fetching during bulk installs.
• Batch actions — Download fewer items, then pause, then continue.

Auto Retries That Spiral

A lot of tools retry on failure. If the retry loop is tight, each 429 triggers another 429, and you end up locked out longer.

1. Stop the queue — Don’t keep running the graph while the remote service is throttling.
2. Increase retry delay — Use a longer wait between attempts, not rapid repeats.
3. Add jitter — Randomize the delay so requests don’t bunch up on the reset second.

Shared Limits Across Tools

Some services limit by IP, not by app. A model manager, a browser, and another device can all be “one client” to the provider.

• Pause other downloads — Stop Steam updates, cloud sync, and other heavy fetches.
• Try a different network — A phone hotspot can confirm whether your home IP is throttled.
• Wait out the window — Some providers escalate the cooldown if you keep hammering.

Fixing 429 ComfyUI Error On Downloads, Managers, And APIs

This section is the hands-on playbook. Use the branch that matches where you see the failure. If you’re seeing 429 comfyui error during a manager download, start there before touching anything else.

If It Happens During Model Or LoRA Downloads

1. Download fewer files — Pick one model at a time, then wait for it to finish cleanly.
2. Turn off previews — Skip example image pulls until the main files are done.
3. Use one source per session — Don’t mix multiple repos at once if they share the same host.
4. Clear failed tasks — Remove partial entries so the manager doesn’t re-try them in a loop.

If the host is a public site, peak hours can be rough. Try again later with slower concurrency and you’ll often get through without changing any settings.

If Downloads Fail From Git Hosts Or Model Repos

Some managers fetch from GitHub releases, model hubs, or CDN-backed pages. Those platforms throttle hard when they see many small requests in a row, like metadata calls plus file chunks.

1. Use direct file links — Prefer one clean file URL over pages that trigger extra redirects and checks.
2. Enable resume — A resumable download avoids restarting from zero and re-hitting the limit.
3. Stagger large files — Start the next download only after the first one completes.

If the provider offers an authenticated token for downloads, use it through the official method. Auth often raises limits and reduces false positives. Don’t try to dodge limits with tricks that break terms.

If A Custom Node Calls A Remote API

API limits usually measure requests per minute, tokens per minute, or both. A single ComfyUI graph can burn through those quotas fast when you batch prompts.

• Reduce batch size — Smaller batches mean fewer calls per minute.
• Cache responses — Reuse results for repeated prompts instead of calling again.
• Queue calls — Send requests one at a time with a fixed gap between them.
• Respect Retry-After — If the API tells you when to retry, follow it exactly.

If The Web UI Keeps Throwing 429

1. Close extra tabs — Multiple open sessions can trigger repeated refresh requests.
2. Disable auto-refresh widgets — Live feeds can poll faster than you notice.
3. Hard refresh once — Reload the page after the cooldown, not during the lockout.

Fast Triage Table

Where It Breaks	Most Likely Cause	First Fix To Try
Manager bulk download	High concurrency on a shared host	Set threads to 1–2 and pause between items
Thumbnail gallery	Burst of preview requests	Disable auto-preview and reload after a wait
API node runs	Quota hit from batches and retries	Lower batch size and add backoff with jitter
Every request, all tools	IP-level cooldown	Stop retries and wait longer before trying again

Preventing 429 Errors So Your Queue Keeps Running

Once you’re unblocked, set things up so the same spike doesn’t bite you again. A few small habits do most of the work.

Slow Down In The Right Place

Don’t slow the whole PC if only one tool is noisy. Instead, limit the piece that’s making outbound requests.

• Cap parallel fetches — Keep concurrent downloads low on public hosts.
• Spread heavy actions — Run installs, previews, and metadata pulls in separate sessions.
• Schedule big pulls — Late night or off-peak hours can reduce throttling on shared services.

Add A Backoff Pattern That Actually Works

Backoff means you retry, but you wait longer each time. Add a small random offset so you don’t collide with other clients.

1. Start with a short wait — Try 2–5 seconds for the first retry.
2. Double the delay — 5, 10, 20, 40 seconds is a common shape.
3. Stop after a limit — After a few failed tries, wait a full window or quit the run.

Keep Your Requests Predictable

Services like steady traffic. Bursts can trigger throttles even when your total is small.

• Debounce UI actions — Don’t click refresh repeatedly; wait for the response.
• Reuse cached files — Store models and previews locally once you have them.
• Log request rates — A simple counter per minute helps you see spikes before they break.

Keep A Local Stash So You Don’t Re-Download

A lot of 429 pain comes from re-fetching the same files. Once you have a model, keep it in a stable local folder and point ComfyUI to it, even if you reinstall nodes.

• Back up models — Store checkpoints, LoRAs, and embeddings outside the app folder.
• Reuse previews — Save example images locally and disable auto-fetch for old items.
• Pin versions — Avoid frequent update checks that pull the same metadata over and over.

When The Limit Is Not Your Fault

Sometimes you’re doing everything right and the service still throttles you. It happens during outages, traffic surges, and provider-side tuning.

Look for hints in the response. A long Retry-After, a message about “quota,” or a lockout that lasts hours can point to account limits instead of short burst limits.

Signs You Hit An Account Or Plan Limit

• Same cooldown every time — The wait is fixed and repeats across days.
• Quota wording in the error — The message mentions a usage tier or daily cap.
• Requests are low — You’re sending few calls, yet you still get 429 quickly.

What To Do In That Case

1. Check provider dashboards — Many APIs show request and token usage by minute.
2. Lower cost per request — Cut prompt size, reduce steps, and avoid duplicate calls.
3. Request a higher limit — Some services raise caps after verification or plan changes.
4. Switch endpoints — If a mirror or alternate host exists, spread the load.

If the 429 started suddenly on a workflow that used to run fine, check the provider’s status page or recent change notes. A temporary throttle during incident response can look like a limit problem on your side.

• Record the timestamp — Note when the first 429 appeared and how long the lockout lasted.
• Save the request URL — Keep the domain and path, not your private tokens.
• Capture headers — If you can see Retry-After or rate limit headers, copy them into your notes.

If you run ComfyUI on a shared server, also check whether other users are hammering the same download sources. One teammate can trigger a cooldown for everyone on that IP.