Most of the time Higgsfield is not down, it is queued. Credit generations run in a priority queue. Unlimited and free generations run on a shared pool, so they wait behind everyone else and slow down during peak hours. If a render that used to take forty seconds now takes four minutes, check whether the Unlimited toggle is on before you assume anything is broken.

The second most common cause is a stale page. A hard refresh is the first fix in Higgsfield's guide, because "Something went wrong" is usually a cached bundle or an uncompleted captcha.

If the free tier is hitting that wall every night, do the paid math once. A referral link carries up to 50% off for three hours, and Starter's 270 credits at $19 is the cheapest way off the shared pool. Check the current offer. We may earn a commission if you subscribe through it.

Read the Tile Before You Refresh

The tile tells you which stage you are in, and each stage takes a different action. That is the whole diagnostic.

What the tile showsWhat it meansWhat to do
Queued, not movingWaiting for a slotHover the tile, hit Cancel, retry. Credits return instantly
ProcessingStarted on the backend, cancel may be goneWait. If it fails, credits return automatically
GeneratingRunning, no cancel optionWait. Nothing you click speeds it up
"Something went wrong"Temporary error, or a captcha not completedHard refresh, complete the captcha, retry

The asymmetry at the top is the part worth memorizing. Cancel works instantly and refunds credits while a job is Queued, and stops working once it moves to Processing. A Queued job is free to abandon. A Generating job has already cost you the wait whether or not you close the tab.

Ctrl+Shift+R on Windows, Cmd+Shift+R in Chrome on Mac, Cmd+Option+R in Safari. A normal refresh reuses the cached page.

The Queue Is the Product

Here is the mechanism, from the help center. Credit generations run in the priority queue at maximum speed. Unlimited generations run in the standard queue. Free generations run on a shared pool. At peak, processing limits adjust dynamically, so Unlimited speed and concurrency vary by the hour.

That produces the counterintuitive result: turning the Unlimited toggle on makes a render slower, deliberately, because you stopped paying credits for it. The toggle appears on the generation panel for models covered by your access. Switch it off and the same job returns to the priority queue and deducts credits.

Unlimited is a budget decision, not a speed feature. Deadline tonight, spend the credits. Twelve prompt variants on a Tuesday, flip Unlimited on and accept the wait.

What Normal Looks Like

Higgsfield's enterprise docs put ordinary jobs at 30 seconds to 10 minutes, with heavier video models at 15 to 20 minutes. A still image and a ten-second Kling clip are not the same compute.

Concurrency is the other limit people hit. The free tier runs one generation at a time. Starter runs two video jobs and four image jobs at once. Higher plans add slots, which is why a $99 plan clears a batch faster than a $19 one: same per-clip speed, more in flight.

When It Is Not You

Sometimes it is a real outage. status.higgsfield.ai shows current state, 90-day uptime, and incident history for the developer console, models API, and webhook delivery.

On the API side, a full queue returns 429 with account_queue_full and a Retry-After header, and nothing is created or charged. Wait, resubmit with the same idempotency key, and you will not bill twice.

Before You Email Support

Do the cheap fixes first: hard refresh, complete any captcha, try incognito or another browser, and disable any VPN, since some auth and payment flows block VPN connections.

If you still need Support, send the generation ID, model name, error text, browser and OS. For API issues include the request_id and X-Correlation-ID header.

The bottom line: slow is a queue class, broken is a stage or a stale page, and a real outage lives on the status page. Check the tile, check the toggle, then check status.higgsfield.ai. In that order, most of these resolve without a ticket.