AI File Upload Limits Compared (2026): ChatGPT vs Claude vs Gemini vs NotebookLM

AI File Upload Limits Compared (2026): ChatGPT vs Claude vs Gemini vs NotebookLM

If you research from social content — X threads, LinkedIn posts, Reddit discussions, newsletters — you eventually hit the same wall no matter which AI tool you use: a file won't upload, a project says it's full, or a chat quietly stops accepting new sources. Every major AI assistant caps how much you can feed it, and the caps differ enough that switching tools without checking first wastes real time.

This guide compares ChatGPT, Claude, Gemini, and NotebookLM's file upload limits as they stand today, verified directly against each vendor's own documentation. It also covers the actual fix: most people hit these caps not because they have too much research, but because they upload social content as dozens of small, separate files instead of a few well-organized ones — the problem PostToSource is built to solve, by converting X threads, LinkedIn posts, Reddit threads, and newsletters into clean text or markdown you can consolidate before hitting any of the ceilings below.

The Limits, Side by Side

Tool	Files per upload	Max file size	Notes
ChatGPT	Up to 80 files per 3 hours (Free: 3 per day)	512 MB per file	Text/documents capped at 2M tokens; CSVs ~50MB; images 20MB
Claude (chat)	Up to 20 files per chat	500 MB per file	PDFs limited to 1,000 pages; visual analysis only for PDFs ≤100 pages
Claude (Projects)	Unlimited files, bounded by context window	30 MB per file	Paid plans get RAG mode that extends effective capacity
Gemini	Up to 10 files per prompt	100 MB per file (2 GB for video)	Video capped at 5 min (free) / 1 hour (Pro/Ultra); audio 10 min / 3 hours
NotebookLM	50 sources per notebook (Free/Standard)	500,000 words or 200 MB per source	No page-count limit; higher tiers raise the per-notebook source cap

All figures checked against each vendor's own help documentation on 2026-08-17. NotebookLM's paid-tier source caps (Plus, Pro, Ultra) are covered in more detail in the NotebookLM source limit guide — Google's public documentation doesn't clearly separate those tiers in one place, so treat any number above the confirmed 50-source free tier with caution until you check your own account's plan page.

ChatGPT: Frequent Small Uploads, Not Huge Files

ChatGPT's ceiling is generous on file size — 512 MB per file — but it throttles frequency instead. Free users get 3 uploads per day; every other plan allows up to 80 files every 3 hours, a rolling window that can tighten during peak traffic. Text and document files are also capped at 2 million tokens each, and storage tops out at 25 GB per user, 100 GB per organization.

For social media research, the frequency cap is the one that bites — extracting a week of LinkedIn posts and Reddit threads one at a time can burn through the free-tier daily allowance before lunch. This is the same problem covered in how to use ChatGPT Projects for social media research: extract with PostToSource first, group posts by topic into single files, and upload the batch instead of the individual pieces.

Claude: Two Different Limits Depending on Where You Upload

Claude splits its limits by context. In a regular chat, you can attach up to 20 files per message, each up to 500 MB. Inside a Claude Project, the per-file cap drops to 30 MB, but the file count becomes effectively unlimited — Claude bounds you by total context window instead, and paid plans get an automatic RAG mode that expands that capacity roughly 10x once you approach the limit.

A chat suits a one-off analysis of a handful of threads; a Project is the better home for an ongoing social content archive, the workflow covered in using Claude Projects for social content research. Because Project files are capped at 30 MB each, exporting social posts as clean text through PostToSource keeps every file well under that ceiling regardless of how much source material it represents. Claude also caps PDFs at 1,000 pages, with visual analysis limited to PDFs of 100 pages or fewer.

Gemini: The Tightest File Count, the Most Generous Media Limits

Gemini allows only 10 files per prompt — the strictest count here — but it's the most permissive on file size and media length. Non-video files can reach 100 MB, videos up to 2 GB. Free users are capped at 5-minute videos and 10-minute audio; Google AI Pro and Ultra subscribers get 1-hour video, 3-hour audio, and a 1-million-token context window that makes the 10-file cap less limiting in practice.

That 10-file ceiling makes Gemini a poor fit for uploading social posts one at a time — you'll run out of slots after ten tweets. Converting a week of X threads or newsletter issues into one or two consolidated text files with PostToSource avoids that entirely.

NotebookLM: Sources, Not Files

NotebookLM counts differently than the other three — it caps sources per notebook, not files per upload. Free and Standard accounts get 50 sources per notebook (across up to 100 notebooks), and each source can hold up to 500,000 words or 200 MB, whichever comes first, with no page-count restriction — generous enough that one consolidated file can absorb dozens of social posts without coming close to it.

This is covered in depth in the NotebookLM source limit guide, including the same batching strategy that applies across every tool here.

The Actual Fix: Stop Uploading One Post at a Time

Across all four tools, the pattern is the same. The limit that trips people up almost never fails because there's too much research — it fails because social content gets uploaded in the smallest possible unit: a single tweet as one file, a single LinkedIn post as another, thirty Reddit comments as thirty separate uploads.

PostToSource converts social post links — X threads, LinkedIn posts, Reddit discussions, Substack and Beehiiv newsletters — into clean markdown or plain text you control the grouping of. Instead of exporting one file per post, export one file per topic: "competitor LinkedIn posts, Q3" or "best growth threads this month." That single file counts as one upload against ChatGPT's 3-hour window, one file against Claude's 20-file chat limit or 30 MB Project cap, one of Gemini's 10 slots, or one source against NotebookLM's 50-source ceiling — regardless of how many original posts it contains. This is the same consolidation idea behind converting social post links into AI knowledge bases: fewer, denser files beat many small ones on every platform.

Choosing a Tool Based on Limits Alone

ChatGPT's 80-files-per-3-hours cadence suits frequent small updates better than Gemini's fixed 10-file ceiling. Claude Projects' effectively unlimited file count suits an archive that grows for months. Gemini's 2 GB video / 3-hour audio ceiling (paid tiers) is the most generous for large media files. NotebookLM's 500,000-word-per-source cap, combined with toggling sources on and off per query, suits large organized text libraries — see the full comparison in NotebookLM vs ChatGPT for content creators. Whichever you choose, the batching principle holds: extract first, consolidate by topic, then upload.

Frequently Asked Questions

Which AI tool has the highest file upload limit?

By file size, ChatGPT and Claude's chat interface are close, at 512 MB and 500 MB per file respectively. By file count, Claude Projects is the most permissive since it isn't capped by a fixed number of files, only by total context. Gemini has the strictest count at 10 files per prompt. NotebookLM doesn't use a per-upload file count at all — it caps total sources per notebook (50 on the free tier).

Do these limits reset over time?

Yes, for the frequency-based ones. ChatGPT's 80-files-per-3-hours cap is a rolling window, so it resets gradually rather than at a fixed time. Free-tier ChatGPT's 3-per-day allowance resets daily. Claude and Gemini's limits are per-chat or per-prompt rather than time-based. NotebookLM's source limit is fixed per notebook and only frees up when you delete a source or start a new notebook.

Can I get around these limits without upgrading to a paid plan?

In most cases, yes. The limits are almost always hit because of upload granularity, not because the free tier is too small for typical research. Consolidating social posts into fewer, topic-grouped files with a tool like PostToSource before uploading reduces file count without reducing the amount of content you actually have access to.

Does NotebookLM's source limit work the same way as a file upload limit?

Not exactly. NotebookLM caps the number of distinct sources per notebook rather than files per single upload action, and each source can be very large — up to 500,000 words. That makes it more forgiving for large single documents but stricter than the others if you're used to uploading many small files, since each one still consumes a source slot.