What "rate limit" actually means on Grok Imagine
Standard API rate limits are simple: N requests per minute, hard reject after N+1. Grok Imagine's consumer product does something more layered.
Every generation consumes quota units. The unit cost varies by render type — a text-to-image at 1024x1024 costs one unit, a 720p video render costs somewhere between 5 and 10 units depending on length. Your account has three concurrent quotas: a rolling 24-hour daily budget (per-tier), a rolling 2-hour burst cap (universal), and a soft peak-hours modifier that throttles heavy users during high-load windows without formally rejecting requests.
Hitting any of the three triggers a different behavior. The 24-hour ceiling returns a hard "daily limit reached" error. The 2-hour burst cap returns a "generating too quickly" backoff. The peak-hours throttle silently degrades performance (longer queues, lower resolution than requested, mid-render failures that still consume quota).
This is why users report inconsistent behavior. "I hit the limit yesterday at 20 videos but today at 8" is usually the 2-hour burst cap firing on a heavier usage pattern, not the 24-hour ceiling dropping.
Per-tier ceilings after the Fall 2026 SuperGrok repricing
xAI restructured SuperGrok in Fall 2026 — Lite tripled from ~$10 to $30, a new Plus tier at $100 landed between Lite and Heavy, and Heavy stayed at $300. Current tier ceilings:
- Free tier — zero image or video generation (both removed March 19, 2026).
- SuperGrok Lite ($30/month) — ~1,500 attempted images and ~240 attempted videos per month. At the ~40% moderation success rate reported by the community, usable output lands around 600 images / 96 videos.
- SuperGrok Plus ($100/month, new) — 3x Lite volumes (~4,500 attempted images / ~720 videos, ~1,800 usable images / ~288 usable videos).
- SuperGrok Heavy ($300/month) — ~10,000 attempted images and ~2,000 attempted videos per month, ~4,000 / ~800 usable at the same success rate. Plus Grok 4.5 access and priority queue.
- X Premium ($8/month via X) and X Premium+ ($40/month via X) — bundled X social-media subscriptions that include some Grok access alongside X features. These sit alongside the SuperGrok tiers rather than replacing them.
Every number is community-reported. xAI does not publish exact per-tier ceilings. Numbers are current as of Fall 2026 after the SuperGrok repricing announcement. For a full breakdown of what each tier costs and what you get, see Grok Imagine Free vs Premium (2026).
The 2-hour rolling burst cap
The 2-hour burst cap is the mechanic most users hit first and understand least. It exists to prevent scripted bulk generation.
Mechanic: within any 2-hour window, you can render approximately 15-30% of your daily quota before the burst cap fires. Once fired, you get a "generating too quickly" or similar backoff, and further requests get rejected until the window rolls.
Rolling behavior: the window is not a fixed 12am-2am / 2am-4am / etc. clock. It rolls per generation — the "oldest" render in your 2-hour history determines when capacity frees up. If you generated 10 renders between 8pm and 8:30pm, capacity restores incrementally from 10pm to 10:30pm.
Practical advice: pace your generations. Two per 10 minutes rather than 10 per 10 minutes will keep you clear of the burst cap even on modest tiers.
Fair-use throttling: silent, undocumented, real
Beyond the two hard ceilings, xAI runs an undocumented "fair use" algorithm that identifies heavy users during peak platform load and adds latency + degradation to their requests.
Observed behaviors during throttled hours:
- Queue times balloon from ~90 seconds to 4-6 minutes.
- Some renders return at lower resolution than requested with no UI notice.
- Some renders partially fail after consuming quota (this is the most-complained-about behavior — you paid for capacity you didn't get).
- Priority reorders push higher-tier subscribers ahead of lower tiers.
The throttle activates during US 6pm-11pm ET weekdays most consistently, plus Sunday nights. Regional users generating on off-peak windows for their local timezone often see substantially different daily throughput than the community-reported averages.
The algorithm is not public and not appealable. Community consensus is that it triggers on aggregate account signal — total renders in the last 24h, the last 2h, and the last hour combined — but exact weights are unknown.
What resets, when
Consolidated reset behavior:
- Daily ceiling — rolling 24-hour window per generation. A render at 3pm Monday frees its slot at 3pm Tuesday.
- Burst cap — rolling 2-hour window per generation. A render at 3:15pm frees its slot at 5:15pm.
- Fair-use throttle — resets based on aggregate account signal decay. Typically a full recovery within 6-12 hours of not generating.
- Peak-hours modifier — time-of-day based, not account based. Clears at ~11pm ET automatically.
What this means for planning: you cannot say "my quota resets at midnight." You can say "the last 24 hours of my generation activity is what governs my current capacity." Grok's reset design is inherently rolling; scheduling around a fixed clock time will misfire.
Options when throttled
When you're throttled and need to keep generating, the practical paths.
Downgrade resolution. Dropping from 720p to 480p roughly doubles your daily video count. Quality is meaningfully lower but the render succeeds.
Wait for the specific window that fired. If burst cap fired, wait 2 hours. If daily ceiling fired, wait for individual slots to roll (partial capacity returns incrementally, not all at once). If peak-hours throttle, wait until after 11pm ET.
Upgrade the tier. Each step up roughly doubles practical throughput. If you're consistently hitting Premium+ throttle at 10-15 videos, SuperGrok Heavy is priced against Runway Enterprise / Midjourney Mega at ~$300/month for professional creative use.
Use a purpose-built alternative for your specific use case. Grok Imagine's throttle is designed for a general-purpose tool serving broad demand. Per-persona LoRA companion tools like Sandra on Sloane sidestep this because the pricing model is credit-per-render (25 credits per 5-sec or 40 credits per 8-sec video, no daily wall — you buy credits, they don't expire, you generate against them). No rolling windows, no fair-use throttle, no peak-hours degradation. Plus at $9.99/month includes 100 credits (4 videos or 20 photos); Premium at $19.99/month includes 350 credits (14 videos or 70 photos); credit packs at $9.99/100, $19.99/250, $29.99/500 for heavier use. The trade is a fixed roster of characters vs. Grok's open prompting.
Which mechanic hits you first, by usage pattern
Bulk generator (many renders in a short window). The 2-hour burst cap fires first. Pace your generations to stay under 15-30% of daily quota per 2 hours.
Sustained heavy user (30+ videos/day at 720p). The daily ceiling fires first on Premium/Premium+; fair-use throttle fires first on Heavy.
Peak-hours user (evening generation on weekdays). The peak-hours throttle fires first regardless of tier. Shifting to off-peak eliminates most of the degradation.
Casual user (few renders per day). None of the mechanics fire. You're operating well within all three windows.
Most frustration comes from users who fit the "peak-hours bulk generator" pattern — they hit all three throttles simultaneously and interpret it as random breakage. It isn't random; it's three overlapping systems firing on the same aggregate signal.
