Gemini Usage Limits: Free, AI Pro & AI Ultra (2026)
Google overhauled Gemini’s usage limits at I/O 2026: flat daily prompt caps are gone, replaced by compute-based limits that refresh every 5 hours under a weekly ceiling. How much you get depends on your plan. Per Google’s help center, Google AI Plus gets 2x the standard (free) allowance, Google AI Pro ($19.99/month) gets 4x, and Google AI Ultra (from $99.99/month) gets 5x or 20x higher than AI Pro depending on tier. Context windows are published outright — 32,000 tokens free, 128,000 on Plus, 1 million on Pro and Ultra — and heavy features like Deep Research, image and video generation, and Deep Think burn compute fastest. Here are the actual numbers as of mid-2026.
The Limits at a Glance (as of 2026)
Free
- Standard compute allowance: refreshes every 5 hours until you hit the weekly cap
- Top-model access is thin — independent trackers put free Gemini Pro-class prompts at only a handful per day before you’re moved to Flash (Zenken’s June 2026 tracker)
- Roughly 20 image generations/day and 5 Deep Research reports/month; no video generation
- 32,000-token context window
Google AI Plus
- 2x standard limits; about 30 top-model prompts and 12 Deep Research runs/day, ~50 images/day, and video via a monthly credit pool
- 128,000-token context window
Google AI Pro ($19.99/month)
- 4x standard limits; roughly 100 top-model prompts/day, 20 Deep Research runs/day, ~100 images/day, and about 3 Veo video clips/day
- 1 million-token context window
Google AI Ultra ($99.99–$199.99/month)
- 5x or 20x AI Pro’s limits depending on tier; trackers log roughly 500 top-model prompts/day, 120–200 Deep Research runs/day, up to ~1,000 images/day, and ~5 video clips/day
- Exclusive Deep Think access (about 10 prompts/day at a 192,000-token window)
Google revises these numbers frequently, and the per-feature figures above are the practical ceilings logged by third-party trackers rather than a table Google prints in one place. Caps also differ across the Gemini app, Workspace, and the API — confirm against Google’s help center before relying on a specific figure.
Compute-Based Limits Replaced Flat Prompt Caps at I/O 2026
Gemini used to publish per-model daily prompt caps (100/day on 2.5 Pro for AI Pro subscribers, 500/day for Ultra). At I/O 2026, Google switched the Gemini app to compute-based metering: every request draws down an allowance based on prompt complexity, the features used, and chat length. Burn through it and you first hit a 5-hour rolling limit; keep going and a weekly cap kicks in.
The plan multipliers are the part Google states plainly: AI Plus 2x, AI Pro 4x the standard allowance, and AI Ultra 5x or 20x above AI Pro. Pro-class and extended-thinking models consume far more compute than Flash, so the old intuition survives: when your allowance runs low, Gemini leans on lighter models, and upgrading buys you dramatically more time on the strongest one. If you hit a wall mid-task, Google now also sells pay-as-you-go credits to keep working without upgrading plans.
Deep Research, Image, and Video Generation Are Metered Separately
Gemini’s heaviest features carry their own trackable ceilings (figures from Zenken and 9to5Google, as of mid-2026):
- Deep Research — the multi-step, cited-report mode — allows about 5 reports/month free, ~12/day on Plus, ~20/day on Pro, and ~120–200/day on Ultra. Each run does a lot of browsing and reasoning, so it’s also one of the fastest ways to spend your compute allowance.
- Image generation runs roughly 20/day free, ~50/day on Plus, ~100/day on Pro, and up to ~1,000/day on Ultra on the Nano Banana models.
- Video generation (Veo) is paid-only: none on free, a monthly credit pool on Plus, and roughly 3 clips/day on Pro and 5/day on Ultra.
Daily caps reset around midnight Pacific time and monthly ones on the 1st, per Zenken’s tracking — so exhausting Deep Research doesn’t stop you from chatting, and vice versa.
The Context Window Is Its Own Kind of Limit
Separate from how much you can generate is how much a single conversation can hold. These numbers Google publishes directly: 32,000 tokens on free (roughly 25–50 pages), 128,000 on AI Plus, and 1 million tokens on AI Pro and Ultra — enough for long documents or a sizable codebase. Ultra’s Deep Think mode runs a smaller 192,000-token window. Remember that under compute-based metering, stuffing the window has a price: huge prompts and long-running chats draw down your allowance faster. If you don’t need the whole document, trimming to the relevant parts keeps things fast and cheap.
How to Stretch Your Gemini Limits
- Reserve Gemini 3.1 Pro for hard reasoning; let Flash handle routine tasks — Pro-class prompts burn compute several times faster.
- Batch Deep Research for questions that genuinely need multi-step browsing; on free you only get about 5 runs a month.
- Trim context to what matters instead of pasting entire documents every time.
- Start heavy sessions after a reset — the compute meter refreshes every 5 hours, so a fresh window beats the tail end of one.
- Upgrade or top up: AI Pro’s 4x allowance (or Ultra’s 5–20x that) is the durable fix, and pay-as-you-go credits cover the occasional spike.
Troubleshooting
How many prompts can I send on free Gemini?
There’s no published flat number anymore — free users get the standard compute allowance (5-hour refresh, weekly cap). In practice, trackers log only a handful of Gemini Pro-class prompts per day on free before you’re shifted to Flash, plus about 20 images/day and 5 Deep Research reports/month.
What does Google AI Pro raise?
AI Pro ($19.99/month) gives 4x the standard allowance — roughly 100 top-model prompts/day and 20 Deep Research runs/day in third-party tracking — plus the 1M-token context window and video generation.
Does Gemini have a separate image generation limit?
Yes. Image generation has its own daily ceiling — roughly 20/day free rising to ~1,000/day on Ultra — though under compute-based metering, heavy generation also drains your overall allowance.
What happens when I hit my Gemini limit?
For text you’re usually switched to a lighter Flash model, and heavy compute use triggers the 5-hour limit first, then the weekly cap. You can wait for the refresh, buy pay-as-you-go credits, or upgrade. Deep Research and media generation pause until their own quotas refresh.
Is Google AI Ultra worth it over AI Pro?
Only for heavy users. Ultra ($99.99–$199.99/month) runs 5x or 20x AI Pro’s allowance, roughly 500 top-model prompts and 120–200 Deep Research runs/day, and is the only tier with Deep Think (~10 prompts/day).
Quick Reference
| Plan | Compute allowance | Top-model prompts (approx.) | Deep Research | Context window |
|---|---|---|---|---|
| Free | Standard (5-hr refresh, weekly cap) | A handful/day | 5/month | 32k tokens |
| AI Plus | 2x standard | ~30/day | ~12/day | 128k tokens |
| AI Pro ($19.99/mo) | 4x standard | ~100/day | ~20/day | 1M tokens |
| AI Ultra ($99.99–$199.99/mo) | 5x–20x AI Pro | ~500/day | ~120–200/day | 1M tokens (192k in Deep Think) |
Multipliers and context windows are Google-published; per-feature counts are third-party-tracked ceilings as of mid-2026 and subject to frequent change.
If you’d rather have an AI that acts on your inbox and calendar than one you prompt in a chat window, an assistant like Carly runs from your email and triggers on real events — see the best AI personal assistants.
Related guides: Claude usage limit · ChatGPT Plus limits · Cursor usage limits · NotebookLM limits · GitHub copilot limits · Grok limits · HubSpot contact limit
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
See what people say
"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.
Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.
On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."
