Skip to content
induwara.lk
Premium
induwara.lkAI · Reference

AI File Upload Limits — ChatGPT, Claude, Gemini & More

Stop guessing why a file is “too large.” Type your file's size, type and page count to see which of 7 AI assistants will accept it — with the exact max size, PDF page limit and files-per-chat for each, every figure linked to its official source.

By Induwara AshinsanaUpdated Jul 9, 2026
Will your file fit?7 AI assistants
Vendor-verified · Jul 2026

= 44 MB

Needed for page-limit checks (Claude, Gemini).

Try
4 of 7 assistants can take this file.Best match: ChatGPT (most headroom).
ChatGPT
OpenAI
Best
Accepted

44 MB · 220 pages within limits

468 MB of headroom

Claude
Anthropic
Rejected

44 MB exceeds Claude's 30 MB limit

Gemini
Google
Accepted

44 MB · 220 pages within limits

6 MB of headroom

Copilot
Microsoft
Accepted

44 MB · 220 pages within limits

6 MB of headroom

Perplexity
Perplexity AI
Rejected

44 MB exceeds Perplexity's 40 MB limit

Grok
xAI
Accepted

44 MB · 220 pages within limits

4 MB of headroom

DeepSeek
DeepSeek
Not published

DeepSeek does not publish a size limit for PDF

Published limits — PDF

AssistantMax sizePDF pagesFiles / chatFree tierSource
ChatGPT
OpenAI
512 MBToken-bound10Free plan: 3 file uploads/day
Claude
Anthropic
30 MB10020Included on free plan (lower message limits)
Gemini
Google
50 MB1,00010Free tier available (48-hour file retention)
Copilot
Microsoft
50 MB20Available on free Copilot
Perplexity
Perplexity AI
40 MB10Free plan: limited daily uploads (40 MB/file)
Grok
xAI
48 MB100Free Grok on X (rate-limited)
DeepSeek
DeepSeek
Not publishedFree app upload (limits not officially published)

“Token-bound” = no fixed page cap; the document is limited by a per-file token budget instead. Free-tier allowances change often.

Every figure is a hand-verified constant from the vendor's own help pages (last checked 9 Jul 2026), each linked in the table. Cells a vendor doesn't publish read “Not published” — never guessed. Vendors change limits without notice; confirm against the linked source before a deadline.

How it works

This is a lookup, not a guess. Each AI assistant publishes its upload caps on its own help pages; this tool stores those figures as verified constants and runs a single deterministic check — exact number comparison, no modelling — to tell you whether a file clears every gate. It never reads your actual file: you type the size, so nothing is uploaded and there is no privacy cost.

  1. Normalise the size to megabytes.The entered value is multiplied by 1 (MB), 1⁄1024 (KB) or 1024 (GB), so every comparison happens in one unit.
  2. Check the format.If an assistant doesn't accept that file type at all — for example Grok rejects XLSX spreadsheets — it's an immediate no.
  3. Check the size cap for that type.Many vendors set a lower ceiling for images (commonly 20 MB) than for documents. If the vendor doesn't publish a cap for that type, the tool returns “Not published” instead of pretending to know.
  4. Check the PDF page limit. Claude caps a PDF at 100 pages (600 at a 1M-token context) and Gemini at 1,000 pages. ChatGPT sets no page limit — its documents are bound by a 2-million-token per-file budget instead.
  5. Flag the token caveat.For token-bound assistants, the tool estimates document tokens at roughly 500 tokens per page (OpenAI tokenizer guidance, labelled as an estimate) and warns when a document approaches the cap — an amber “accepted with caveat,” never a hard rejection unless the published token cap is actually exceeded.

The recommendation picks the accepted assistant with the greatest size headroom (its cap minus your file), breaking ties by the higher PDF page limit. Every verdict is computed two independent ways in code — a short-circuit pass and a full-condition re-check — and the two must agree, so a logic slip can't silently ship a wrong answer.

Figures cover the consumer apps and public APIs of 7assistants and were reconciled to each vendor's official documentation on 2026-07-09.

Worked examples

44 MB, 220-page client PDF

  1. Normalised size: 44 MB.
  2. ChatGPT: 44 ≤ 512 MB, and 220 pages ≈ 110,000 tokens ≪ 2M cap → Accepted (468 MB headroom).
  3. Claude: 44 MB > 30 MB per-file cap → Rejected on size (it also exceeds the 100-page limit).
  4. Gemini: 44 ≤ 50 MB and 220 ≤ 1,000 pages → Accepted (6 MB headroom).
  5. Perplexity: 44 MB > 40 MB → Rejected.
  6. Verdict: 4 of 7 accept; best match ChatGPT — it clears both gates with the most room to spare.

18 MB photo (PNG)

  1. Type: image, size 18 MB, no page count.
  2. ChatGPT: 18 ≤ 20 MB image cap → Accepted (2 MB headroom).
  3. Gemini: 18 ≤ 20 MB inline cap → Accepted (2 MB headroom).
  4. Copilot: 18 ≤ 50 MB → Accepted (32 MB headroom).
  5. Verdict: 6 of 7 accept; best match Copilot for the largest headroom. DeepSeek stays 'Not published'.

21 MB photo — the boundary case

  1. Type: image, size 21 MB.
  2. ChatGPT: 21 > 20 MB → Rejected: '21 MB exceeds ChatGPT's 20 MB limit'.
  3. Gemini: 21 > 20 MB → Rejected for the same reason.
  4. Claude (30 MB), Copilot (50 MB), Perplexity (40 MB), Grok (48 MB): all still Accepted.
  5. Lesson: a single megabyte over the shared 20 MB image ceiling flips ChatGPT and Gemini to a reject — resize before uploading.

Frequently asked questions

Sources & references

Every figure on this page was cross-checked against the vendor pages above on 2026-07-09. Where a vendor does not publish a clear number, the cell reads “Not published” rather than an estimate. AI providers change these limits frequently; the page is reviewed on a rolling basis and each figure links to its live source.

Related tools

Rate this tool
Be the first to rate

Comments & feedback

Spotted a bug or want an improvement? Tell us — our team reviews every comment, and good ideas get built. Comments are public and anonymous.

Spotted a limit that's changed, or an assistant we should add?

Email me at [email protected] — most fixes ship within 24 hours.