Last updated: 2026-06-29

QR Code Data Capacity

QR codes encode data in a two-dimensional matrix of dark and light modules. How many characters a QR code can hold depends on three independent parameters: its version (which sets the physical grid size), its error correction level (which trades capacity for damage resilience), and the encoding mode used for the payload data.

What QR versions are

The QR code standard (ISO/IEC 18004) defines 40 versions. A version-1 code measures 21×21 modules; each subsequent version adds four modules per side, so a version-$v$ code is $(17 + 4v)$ modules square. The practical consequence is physical size: when printed at a fixed module size, a version-40 code is $\frac{177}{21} \approx 8.4$ times wider (and taller) than a version-1 code.

Higher versions increase both capacity and physical area. Scanners can reliably decode a code whose modules are at least about 0.3 mm, so the maximum useful version for a small sticker is much lower than for a billboard or screen-displayed QR code.

Error correction levels

ISO/IEC 18004 defines four error correction levels, implemented with Reed–Solomon coding:

"Recovery capability" means the fraction of the code's total module area that can be damaged, obscured, or misprinted while the scanner can still reconstruct the full payload. The Reed–Solomon parity codewords that enable this recovery occupy space that would otherwise hold data, so every step up in error correction level reduces usable capacity.

For a URL on a screen — scanned under ideal lighting, unobstructed — level L or M is almost always sufficient. Level H is appropriate when the code appears on a physical surface subject to wear: industrial labels, product packaging, garments, or outdoor signs.

Data encoding modes

The QR standard supports four encoding modes. The choice of mode determines how many bits each character consumes:

Numeric (digits 0–9 only): 10 bits per three digits on average. The most compact option for digit-only data such as long serial numbers or phone numbers.

Alphanumeric (digits, uppercase A–Z, space, and nine punctuation marks — dollar, percent, asterisk, plus, hyphen, period, slash, and colon): 11 bits per two characters. Efficient for short codes, tracking identifiers, and URLs that happen to fit within the uppercase-only alphabet.

Byte (any 8-bit value): 8 bits per byte. Handles any text encoded as UTF-8, including lowercase, mixed case, and non-ASCII characters. A standard HTTPS URL like https://example.com/path?q=1 requires byte mode.

Kanji (Shift-JIS double-byte characters): 13 bits per character. Slightly denser than byte mode for common Japanese kanji and kana, but requires the scanning application to support kanji mode decoding — not all do.

Worked example

A QR code is needed to encode the URL https://example.com/register (33 bytes of ASCII). The designer wants it to be scannable even if a corner is worn off, so they choose error correction level M. What version is the minimum that fits?

At EC-M in byte mode, the capacities by version are:

Version 3 (33×33 modules) is the smallest that fits. Switching to level L does not reduce the required version: at EC-L, version 2 holds only 32 bytes — still below 33 — so version 3 remains the minimum regardless of error correction level chosen.

Choosing version and level in practice

Most QR-code generation libraries pick the minimum version automatically once you specify a level. The relevant design choice is therefore the error correction level:

• Use L for digital displays and single-use printed labels in controlled conditions.
• Use M as the general-purpose default for printed material without unusual damage risk.
• Use Q when the code appears on a product that will be handled (packaging, wristbands).
• Use H for demanding physical environments: outdoor signage, industrial tags, fabric printing.

Avoid unnecessarily high levels for data-dense payloads — every level above L roughly halves the effective capacity, requiring a larger version and a physically bigger symbol.