QR code versions 1 to 40: grid sizes and how much data fits

A QR code's version is its grid size: version 1 is 21×21 modules, each version adds 4 modules per side, and version 40 is 177×177. You don't pick it. The encoder picks the smallest version that holds your content at the error-correction level you chose, so more data means a bigger version and smaller modules on the print.

What decides the version

Three things: how much you encode, which characters it uses, and the error-correction level. The QR standard (ISO/IEC 18004) stores data in one of several modes. The encoder uses the densest mode your whole content allows:

One lowercase letter or a ? switches the whole content to byte mode, which needs about 45% more room per character than alphanumeric (8 bits instead of 5.5). There is also a kanji mode for Japanese text (up to 1,817 characters); Qex doesn't use it.

Error correction adds redundancy so a scratched, smudged or partly covered code still scans. Each level recovers a share of damaged data: L 7%, M 15%, Q 25%, H 30%. Higher levels leave less room for content. A logo in the middle of the code covers modules, so codes with a logo use H.

Real content measured: which version it needs

Each row is encoded with the same encoder Qex uses for every download, at M without a logo and H with one.

The first two rows are the same link. In capitals, HTTPS://QEX.ME/AB3XK fits alphanumeric mode and version 1; in lowercase, https://qex.me/ab3xk needs byte mode and version 2. Every Qex dynamic code holds a short link like the first row, so without a logo it stays version 1 whatever the destination. The rest are static codes, where the pattern holds the whole content. See static vs dynamic QR codes.

All 40 versions: size and capacity

At error correction M, the level Qex uses for codes without a logo. The last column is the width of one module, in millimetres, when the whole code, including its four-module clear margin, is printed 30 mm wide.

Module widths are calculated, not tested minimums. To size a print, use the print-size comparison and the print checklist.

Capacity at each error-correction level

Bytes of text per version (byte mode). For links and text with lowercase letters, this is the column that applies.

Full tables for the other levels, every mode (level M is the table above):

How to get a smaller version

1. Encode a short link instead of the full address. A dynamic code holds a short redirect link, so a long destination with tracking tags still makes a small code. You can also change where it points after printing.
2. Drop tracking tags such as ?utm_source=… from static links. In the table above they take the menu link from version 3 to version 5.
3. Write short links in capitals when the server accepts them. Domain names are not case-sensitive; paths often are, so test first.
4. Skip the logo if size matters: it moves the code from M to H.

Why the smallest version is worth having: the smallest QR code, and how to print it.

Questions

What is the most data a QR code can hold?

Version 40 at error correction L: 7,089 digits, 4,296 alphanumeric characters or 2,953 bytes. A byte is one character of plain English text; accented letters take two bytes and emoji four. A code that dense needs a large print to scan.

What is a version 40 QR code?

The largest QR code: 177×177 modules, holding up to 2,953 bytes at error correction L or 2,331 at M. Grid width is 17 + 4 × version, so 177×177 is version 40, 21×21 is version 1 and 25×25 is version 2. Printed 30 mm wide, a version 40 module is 0.16 mm, so it needs a large print and a close, steady camera.

How many characters fit in a version 4 QR code?

Version 4 (33×33) holds up to 62 bytes of text at error correction M, 78 at L and 34 at H. Capital letters and digits go further: 90 at M, and exactly 50 even at H.

How much fits in a version 1 QR code?

At error correction M: 34 digits, 20 capital letters and digits, or 14 bytes. That is why Qex writes its short links in capitals: HTTPS://QEX.ME/AB3XK is 20 characters, exactly the alphanumeric limit.

Can I choose the version?

Some encoders let you force a larger version, but none can fit data into a smaller one. The encoder picks the smallest version that holds your content at your error-correction level. To get a smaller code, put less in it.