NotebookLM Limits 2026: Every Cap, Quota & Performance Benchmark — The Complete Spec Sheet
Every NotebookLM limit, cap, and generation time in one place — verified after the May Ultra reshuffle and the June Gemini 3.5 update. Source caps, file-size limits, daily quotas, generation-speed benchmarks, and tier-by-tier comparisons. No guessing, no stale pricing. Updated monthly as Google changes the limits.
NotebookLM’s core 2026 limits: 50–600 sources per notebook depending on plan, a fixed 500,000 words / 200 MB per source on every tier, 50–5,000 daily chat queries, and a 1-million-token context window shared across all plans. Free allows 100 notebooks; Ultra tops out at 600 sources and 5,000 daily chats.
TL;DR — Every NotebookLM limit in one place. Source caps (50/100/300/500/600), 500K words / 200MB per source, daily chats (50–5,000), generation times for slides (60–90 sec), audio (3–8 min), video (5–10 min). Free vs Plus ($7.99) vs Pro ($19.99) vs Ultra ($99.99/$200). Updated June 2026.
Updated June 2026. Maintained by a small team of AI super-users who teach multi-AI research and study workflows to researchers, students, and professionals — no affiliate relationships. About this guide →Changelog
Sources per notebook: 50 (Free), 100 (Plus), 300 (Pro), 500 (Ultra 20TB), 600 (Ultra 30TB). Max per source: 500,000 words or 200 MB, all plans. Daily chats: 50 (Free) → 200 (Plus) → 500 (Pro) → 2,500/5,000 (Ultra). Slide deck: 60–90s typical, up to 5 min dense. Audio Overview: 3–8 min; 3/day free, 200/day Ultra. Chat prompt: 10,000 characters. Context window: 1M tokens (all plans). Pricing: Free $0 · Plus $7.99 · Pro $19.99 · Ultra $99.99/$200.
Complete Tier Comparison: Free vs Plus vs Pro vs Ultra
Every limit, side by side. Prices and caps verified June 2026 (after the May 19 Ultra split).
| Limit | Free · $0 | Plus · $7.99 | Pro · $19.99 | Ultra 20TB · $99.99 | Ultra 30TB · $200 |
|---|---|---|---|---|---|
| Notebooks | 100 | 200 | 500 | 500 | 500 |
| Sources per notebook | 50 | 100 | 300 | 500 | 600 |
| Max words per source | 500,000 words (all plans) | ||||
| Max file size per source | 200 MB (all plans) | ||||
| Daily chat queries | 50 | 200 | 500 | 2,500 | 5,000 |
| Chat prompt limit | 10,000 characters (all plans) | ||||
| Context window | 1 million tokens (all plans since Jan 2026) | ||||
| Audio Overviews / day | 3 | 6 | 20 | 100 | 200 |
| Video Overviews / day | 3 | 6 | 20 | 100 | 200 |
| Cinematic Video / day | — | — | 2 | 10 | 20 |
| Reports / day | 10 | 20 | 100 | 500 | 1,000 |
| Flashcards · Quizzes · Mind Maps / day * | 10 | 20 | 100 | 500 | 1,000 |
| Deep Research | 10 / mo | 3 / day | 20 / day | 75 / day | 200 / day |
| Slide Decks & Infographics | Limited | Higher | Higher | Highest | Highest |
| Watermark-free exports | No | No | No | Yes | Yes |
| Agentic features (code exec · new exports) | Rolling out | Rolling out | Rolling out | Yes | Yes |
| Tone & response control · analytics | No | Yes | Yes | Yes | Yes |
What Each Tier Actually Costs (After the May 2026 Reshuffle)
You can’t buy NotebookLM alone — every paid tier is bundled into a Google subscription
This is where most “limits” pages have gone stale. On May 19, 2026, Google reorganized its AI subscriptions and split AI Ultra into two SKUs, giving NotebookLM Ultra two distinct ceilings. The old “Plus = $19.99” or “Plus ≈ $14” framing is wrong now. The correct 2026 price ladder:
Free (Standard) — $0. Any Google account, no card, no expiry. 100 notebooks, 50 sources each, 50 daily chats.
Plus — $7.99/mo via Google AI Plus. Roughly doubles every Free limit, adds notebook sharing, bundles 200 GB storage.
Pro — $19.99/mo via Google AI Pro. The tier most paid users want: 500 notebooks, 300 sources, 500 daily chats, 20 Deep Research/day, Gemini 3 access, 2 TB storage.
Ultra 20TB — $99.99/mo and Ultra 30TB — $200/mo via Google AI Ultra. The 20TB plan delivers 500 sources/notebook and 2,500 daily chats; the 30TB plan pushes to 600 sources and 5,000 daily chats, plus Cinematic Video at scale.
What Changed in 2026 — and the New Kind of Limit
The June 8, 2026 Gemini 3.5 + Antigravity update introduced a limit that isn’t a number
If you’re checking limits to decide what to buy, this matters as much as any cap above. For the first time, your tier doesn’t just control how much you can do — it controls which features exist for you yet. The 2026 timeline:
December 2025 — NotebookLM moved to Gemini 3, replacing the Gemini 2.5 Flash backbone.
January 2026 — the full 1-million-token context window rolled out to all plans, including Free. Context capacity stopped being a paid differentiator.
May 19, 2026 (Google I/O) — Gemini 3.5 launched and AI Ultra split into $99.99 (20TB) and $200 (30TB) SKUs.
June 8, 2026 — NotebookLM moved to Gemini 3.5 + Antigravity, gained a per-notebook secure cloud computer for code execution, 100+ software skills, agentic source discovery (start from scratch), and new export formats.
What the June update unlocked: code execution (each notebook now runs code for real data analysis, not just description); new downloadable formats (charts as .png/.svg, PDF, .docx, Markdown, .xlsx, .pptx, .csv/.json, plus images via Nano Banana); agentic source discovery (start from loose ideas and let NotebookLM find sources via Google Search); and a visible thinking process. In Google’s own evaluations the upgrade posted a 65%+ average win rate over the prior version, including 78.2% on web research and source discovery.
Source Limits: What Can You Upload?
Per-source caps apply identically across all pricing tiers
| Limit Type | Value | What Happens If Exceeded |
|---|---|---|
| Maximum words per source | 500,000 words | Upload fails. Split the document before uploading. |
| Maximum file size | 200 MB | Upload fails. Compress or split the file. |
| Sources per notebook (Free) | 50 | Cannot add more. Delete sources or upgrade. |
| Sources per notebook (Pro) | 300 | Cannot add more. Use multiple notebooks. |
| Sources per notebook (Ultra) | 600 | Cannot add more. Use multiple notebooks. |
| Notebooks per account (Free) | 100 | Delete notebooks to create new ones. |
| Copy-protected PDFs | Not supported | Upload fails silently. Convert to unprotected PDF first. |
| Paywalled web URLs | Not supported | NotebookLM gets the paywall page, not the content. |
Generation Time Benchmarks — How Long Does Each Feature Take?
Based on hands-on testing with notebooks of varying sizes, June 2026
Slide Deck
Slides render progressively. You can continue working. PPTX export adds ~10 seconds. Full slide-speed benchmarks →
Audio Overview
Short focus topics: 3–4 min. Deep dives from 20+ sources: 5–8 min. 80+ languages.
Video Overview
Standard video available on all plans. Cinematic Video (Ultra only) takes longer — fluid animations and rich visuals.
Mind Map
Interactive. Click nodes to explore deeper connections. Fastest Studio tool.
Reports
Study guides, timelines, FAQs, briefing docs. Use suggested formats for best results.
Flashcards & Quizzes
Progress now saved across sessions. Export flashcards to Anki. Delete individual cards.
Data Table
One-click Google Sheets export. Best with structured source content.
Infographic
Sketch Note, Kawaii, Professional, Scientific, Anime, Clay, Editorial, Instructional, Bento Grid, Bricks.
Daily Usage Caps and Reset Behavior
Caps reset on a rolling 24-hour window from first use — not at midnight. No visible counter.
| Feature | Free | Plus | Pro | Ultra 20TB | Ultra 30TB |
|---|---|---|---|---|---|
| Chat queries / day | 50 | 200 | 500 | 2,500 | 5,000 |
| Audio Overviews / day | 3 | 6 | 20 | 100 | 200 |
| Video Overviews / day | 3 | 6 | 20 | 100 | 200 |
| Cinematic Video / day | — | — | 2 | 10 | 20 |
| Reports / day | 10 | 20 | 100 | 500 | 1,000 |
| Flashcards · Quizzes · Mind Maps / day * | 10 | 20 | 100 | 500 | 1,000 |
| Deep Research | 10 / mo | 3 / day | 20 / day | 75 / day | 200 / day |
What counts as a "chat query": Every message you send in the chat panel counts as one query — including follow-up questions, summary requests, and source comparisons. Multi-part questions in a single message count as one query.
What burns through queries fastest: Research sessions with rapid follow-up questions, A/B source comparisons, and iterative refinement prompts. Group related questions into single, detailed prompts to conserve quota — NotebookLM retains context within each conversation.
No visible counter. NotebookLM does not show how many queries you have remaining. You'll hit a "daily limit reached" message when the cap is reached. Plan accordingly for heavy research days.
Supported Source Formats and Per-Format Notes
9 source types supported — each with specific quality considerations
NotebookLM Enterprise & Workspace Limits
Higher caps with governance — for teams that need control, not just capacity
| Edition | Price | Caps | Governance |
|---|---|---|---|
| Workspace Business Standard | $14/user/mo | ~Plus level (200 notebooks · 100 sources) | Core Workspace service; data not used for training |
| Workspace + AI Ultra access | Add-on | Ultra-level caps; early agentic features | Admin-managed, enterprise data protection |
| NotebookLM Enterprise | ~$9/license/mo | Pro-equivalent caps | VPC-SC · IAM · audit logging (Google Cloud) |
The Enterprise difference is governance, not raw numbers. Consumer Ultra gives you the highest caps; Enterprise gives Pro-level caps wrapped in the controls organizations need — VPC Service Controls, IAM-based access, and audit trails. Education Plus and the Teaching & Learning add-on received expanded NotebookLM caps in a gradual rollout starting April 2026. Workspace and Enterprise data is never used to train models and is not reviewed by humans even when feedback is submitted.
Admins control availability: NotebookLM is on by default and can be disabled at the domain, OU, or group level. Certain features remain restricted to users aged 18+.
Hidden Limits & Gotchas — 5 Walls No Plan Removes
Some constraints are architectural, not quota-based. No subscription lifts these.
1. The per-source ceiling is fixed. 500,000 words / 200 MB is the same on every tier, Ultra included. Big scanned books and image-heavy PDFs won’t import regardless of plan — upgrading buys more sources, not bigger ones.
2. There’s no offline mode. NotebookLM runs only in the browser and needs a connection on every plan. Sources are processed on Google’s servers.
3. You’re locked to one model family. NotebookLM runs on Gemini only — no bring-your-own-key, no routing to Claude, GPT, or a local model.
4. Notebooks are isolated. There’s no native cross-notebook query. Research split across 30 themed notebooks can’t be reasoned over in one chat — you consolidate manually. (Workaround: the Claude × NLM Command Center queries across notebooks via MCP.)
5. Export and recovery are thin. Deleted notes can’t be recovered, notebooks can’t be duplicated, and structured citation export is limited. Keep original PDFs, notes, and citations outside NotebookLM if the work must survive a migration.
Two more quiet traps: copy-protected PDFs fail silently (strip DRM first), and paywalled or login-gated URLs import only the gate page, not the content behind it.
Workarounds for Common Limit Issues
Practical solutions when you hit a wall
Hit the 50-source cap? Split your research into multiple themed notebooks. Create a synthesis notebook with notes summarizing findings from each sub-notebook. Use the Claude × NLM Command Center (MCP) to query across notebooks simultaneously.
Document exceeds 500K words? Split into logical sections (e.g., chapters, sections, time periods) and upload each as a separate source. Name descriptively: "Textbook-Ch1-Introduction.pdf" not "part1.pdf."
Hit the daily chat cap? Group related questions into single, detailed prompts. Use multi-part questions in one message. Save insights as notes — notes persist and can be promoted to sources for future queries.
Audio Overview cap reached? Reserve your 3 daily overviews for the densest material. Use text chat for simpler documents where you can read the answers. Consider upgrading to Plus ($7.99/mo) for 6/day, or Pro ($19.99/mo) for 20/day.
Need more than 600 sources? No plan supports more than 600. Use multiple notebooks organized by theme, with a master synthesis notebook. Merge related PDFs before uploading to reduce source count (but be aware this degrades citation specificity).
Slide deck quality issues? Fewer, higher-quality sources produce better slides than many low-quality ones. Add a "Deck Brief" note as a source with your outline and audience description. See our Slide Deck Guide for optimization techniques.
Which Plan Should You Actually Buy?
Pick by the limit you hit most, not the price
Stay on Free if you run one or two projects, generate the occasional Audio Overview, and rarely brush the 50-chat ceiling. The free tier is genuinely powerful — 1M-token context, all core features, no expiry.
Get Plus ($7.99) the first time daily caps shape your behavior: rationing chats, producing audio more than three times a week, or wanting to share a notebook. The lowest-friction upgrade Google offers.
Get Pro ($19.99) when NotebookLM is part of your daily work. The unlock that justifies it is Deep Research at 20/day — the first tier where it behaves like an always-on research agent — plus 300 sources and 500 chats.
Get Ultra ($99.99 / $200) only if you need watermark-free exports, the highest generation volumes, Cinematic Video at scale, or early access to the June agentic features. Overkill for most individuals.
Students: take the $9.99 Pro deal — Pro caps at half price for a year.
Knowing the exact limits turns frustrating failures into predictable workflows — work with the system, not against it
- Every crashed generation has a cause. Source count, word limits, daily caps — when you know the boundaries, you design around them instead of hitting walls.
- Performance benchmarks save hours. Know exactly how long slide decks, audio overviews, and quiz generation take for your source sizes — stop guessing.
- Workarounds for every limit. The spec sheet includes tested strategies for exceeding soft limits — chunking, notebook splitting, priority queuing.
Full benchmark data and workaround library below ↓
Frequently Asked Questions
How many sources can I add to a NotebookLM notebook?
Sources per notebook scale by tier: 50 (Free), 100 (Plus), 300 (Pro), 500 (Ultra 20TB), and 600 (Ultra 30TB). Each source can be up to 500,000 words or 200 MB, whichever comes first. Sharing a notebook does not raise a collaborator’s cap. Supported formats: PDF, Google Docs, Slides, Sheets, YouTube, audio, web URLs, images, EPUB, Word (.docx), and pasted text.
How much does NotebookLM cost in 2026?
NotebookLM is free (Standard, $0). Paid tiers ship inside Google AI subscriptions: Plus at $7.99/month via Google AI Plus, Pro at $19.99/month via Google AI Pro, and Ultra at $99.99 (20TB) or $200/month (30TB) via Google AI Ultra after the May 19, 2026 reshuffle. US students pay $9.99/month. You cannot buy NotebookLM as a standalone product.
What is the maximum file size for NotebookLM sources?
500,000 words OR 200 MB per source, whichever is reached first. There is no page limit. This ceiling is identical on every plan, Ultra included — upgrading buys you more sources, not bigger ones. Copy-protected PDFs fail during upload; split documents over 500K words before uploading.
What is the NotebookLM daily chat limit?
Daily chat queries are 50 (Free), 200 (Plus), 500 (Pro), 2,500 (Ultra 20TB), and 5,000 (Ultra 30TB). Each message counts as one query including follow-ups; multi-part questions in one message count as one. Caps reset on a rolling 24-hour window from first use, not at midnight, and there is no visible counter.
How long does slide deck generation take?
Typically 60–90 seconds for notebooks with fewer than 10 sources, 2–5 minutes for 10–30 sources, and up to 5 minutes for dense notebooks (30+ large PDFs). Slides render progressively and you can keep working. See our Slide Deck Guide for optimization.
How long does Audio Overview generation take?
Audio Overviews generate in 3–8 minutes depending on volume and complexity. Short focused overviews finish in 3–4 minutes; deep dives from 20+ sources take 5–8 minutes. Free gets 3 per day, Ultra up to 200 per day.
Is the context window different on paid plans?
No. NotebookLM uses Gemini’s full 1-million-token context window on every plan, including Free, since January 2026. Combine it with Claude + Gemini → Plans differ in daily caps and source counts, not context capacity. A single chat prompt is capped at 10,000 characters.
What is the NotebookLM Deep Research limit?
Deep Research is 10 reports per month on Free, then 3 per day on Plus, 20 per day on Pro, 75 per day on Ultra 20TB, and 200 per day on Ultra 30TB. Pro is the first tier where Deep Research works as an always-on research agent.
Did NotebookLM limits change in June 2026?
The numeric caps did not change on June 8, 2026, but the model and feature set did. NotebookLM moved to Gemini 3.5 plus the Antigravity framework, added a per-notebook secure cloud computer for code execution, 100+ software skills, agentic source discovery, and new downloadable formats (PDF, DOCX, XLSX, PPTX, CSV/JSON, charts). These agentic features rolled out to Google AI Ultra and Workspace business customers first, reaching other tiers over time.
What are the NotebookLM Enterprise limits?
NotebookLM Enterprise (via Google Cloud, about $9/license/month) provides Pro-equivalent caps plus VPC-SC, IAM, and audit logging. Workspace Business Standard ($14/user/month) includes NotebookLM at roughly Plus-level caps with enterprise data protection. Workspace and Enterprise data is never used to train models.
Do generation limits carry over to the next day?
No. Daily caps reset every 24 hours on a rolling window and unused quota does not carry over. There is no visible counter — you hit a limit message when the cap is reached.
Does Google use my data to train AI models?
No. Google states that personal data and content uploaded to NotebookLM are not used to train foundation models. Workspace and Enterprise editions add VPC-SC compliance and audit trails, and their data is not reviewed by humans even when feedback is submitted.