QR Code Capacity Reference: Versions, Modes, and Character Limits

March 30, 2026

How much data can a QR code hold? The answer depends on three variables: the version (size of the code), the encoding mode (type of data), and the error correction level (redundancy). The tables below cover all 40 versions across all 4 modes and 4 error correction levels, drawn from ISO/IEC 18004:2015 -- the international standard that defines QR code structure and encoding.

Contents

Versions and module counts
Encoding modes
Capacity tables
Symbol structure
How version is chosen
Practical limits

Versions and module counts

QR code "versions" refer to symbol sizes, not software versions. Version 1 is 21×21 modules; each step adds 4 modules per side. Version 40 is 177×177 modules. The formula is: modules per side = 4 × version + 17.

Version	Size (modules)	Total modules	Max data (bytes, EC L)
1	21 × 21	441	19
2	25 × 25	625	34
3	29 × 29	841	55
4	33 × 33	1,089	80
5	37 × 37	1,369	108
7	45 × 45	2,025	154
10	57 × 57	3,249	271
15	77 × 77	5,929	520
20	97 × 97	9,409	858
25	117 × 117	13,689	1,273
30	137 × 137	18,769	1,732
35	157 × 157	24,649	2,303
40	177 × 177	31,329	2,953

Not all modules store data. Finder patterns, timing patterns, format information, version information (v7+), and alignment patterns (v2+) occupy fixed regions. The remaining modules carry interleaved data codewords and Reed-Solomon error correction codewords.

Encoding modes

The encoder automatically picks the most efficient mode for your input. Numeric is most efficient; byte is the fallback for general text. Understanding the modes explains why a number-only string produces a simpler QR code than the same length of mixed characters.

Mode	Character set	Bits per char	Notes
Numeric	0–9 only	3.33 avg (groups of 3 digits in 10 bits)	~3x more efficient than byte mode
Alphanumeric	0–9, A–Z (uppercase only), space, `$ % * + - . / :`	5.5 avg (pairs of chars in 11 bits)	~1.8x more efficient than byte; lowercase letters are NOT included
Byte	ISO 8859-1; in practice UTF-8 is widely accepted	8 (one byte per character)	Baseline; multi-byte UTF-8 chars (emoji, CJK) consume 2–4 bytes each
Kanji	Shift JIS double-byte characters	13 per Shift JIS character	~1.2x more efficient than byte for Japanese/CJK content

Why uppercase URLs produce smaller QR codes: A URL containing only digits, uppercase letters, and $%*+-./: qualifies for alphanumeric mode (~1.8x more efficient than byte). Lowercase forces byte mode. This is why some URL shorteners use all-caps slugs -- it visibly reduces QR code complexity.

Capacity tables

Maximum characters per version, error correction level, and encoding mode. Source: ISO/IEC 18004:2015, Table 1.

L — up to 7% recovery M — up to 15% recovery Q — up to 25% recovery H — up to 30% recovery

Ver	L	M	Q	H
1	41	34	27	17
2	77	63	48	34
3	127	101	77	58
4	187	149	111	82
5	255	202	144	106
6	322	255	178	139
7	370	293	207	154
8	461	365	259	202
9	552	432	312	235
10	652	513	364	288
11	772	604	427	331
12	883	691	489	374
13	1022	796	580	427
14	1101	871	621	468
15	1250	991	703	530
16	1408	1082	775	602
17	1548	1212	876	674
18	1725	1346	948	746
19	1903	1500	1063	813
20	2061	1600	1159	919
21	2232	1708	1224	969
22	2409	1872	1358	1056
23	2620	2059	1468	1108
24	2812	2188	1588	1228
25	3057	2395	1718	1286
26	3283	2544	1804	1425
27	3514	2701	1933	1501
28	3669	2857	2085	1581
29	3909	3035	2181	1677
30	4158	3289	2358	1782
31	4417	3486	2473	1897
32	4686	3693	2670	2022
33	4965	3909	2805	2157
34	5253	4134	2949	2301
35	5529	4343	3081	2361
36	5836	4588	3244	2524
37	6153	4775	3417	2625
38	6479	5039	3599	2735
39	6743	5313	3791	2927
40	7089	5596	3993	3057

Numeric mode: digits 0–9 only.

Ver	L	M	Q	H
1	25	20	16	10
2	47	38	29	20
3	77	61	47	35
4	114	90	67	50
5	154	122	87	64
6	195	154	108	84
7	224	178	125	93
8	279	221	157	122
9	335	262	189	154
10	395	311	221	174
11	468	366	259	200
12	535	419	296	227
13	619	483	352	259
14	667	528	376	283
15	758	600	426	321
16	854	656	470	365
17	938	734	531	408
18	1046	816	574	452
19	1153	909	644	493
20	1249	970	702	557
21	1352	1035	742	587
22	1460	1134	823	640
23	1588	1248	890	672
24	1704	1326	963	744
25	1853	1451	1041	779
26	1990	1542	1094	864
27	2132	1637	1172	910
28	2223	1732	1263	958
29	2369	1839	1322	1016
30	2520	1994	1429	1080
31	2677	2113	1499	1150
32	2840	2238	1618	1226
33	3009	2369	1700	1307
34	3183	2506	1787	1394
35	3351	2632	1867	1431
36	3537	2780	1966	1530
37	3729	2894	2071	1591
38	3927	3054	2181	1658
39	4087	3220	2298	1774
40	4296	3391	2420	1852

Alphanumeric mode: 0–9, A–Z (uppercase only), $%*+-./: and space.

Ver	L	M	Q	H
1	17	14	11	7
2	32	26	20	14
3	53	42	32	24
4	78	62	46	34
5	106	84	60	44
6	134	106	74	58
7	154	122	86	64
8	192	152	108	84
9	230	180	130	98
10	271	213	151	119
11	321	251	177	137
12	367	287	203	155
13	425	331	241	177
14	458	362	258	194
15	520	412	292	220
16	586	450	322	250
17	644	504	364	280
18	718	560	394	310
19	792	624	442	338
20	858	666	482	382
21	929	711	509	403
22	1003	779	565	439
23	1091	857	611	461
24	1171	911	661	511
25	1273	997	715	535
26	1367	1059	751	593
27	1465	1125	805	625
28	1528	1190	868	658
29	1628	1264	908	698
30	1732	1370	982	742
31	1840	1452	1030	790
32	1952	1538	1112	842
33	2068	1628	1168	898
34	2188	1722	1228	958
35	2303	1809	1283	983
36	2431	1911	1351	1051
37	2563	1989	1423	1093
38	2699	2099	1499	1139
39	2809	2213	1579	1219
40	2953	2331	1663	1273

Byte mode: ISO 8859-1 / UTF-8. Multi-byte UTF-8 characters (emoji, most CJK) consume 2–4 bytes each.

Ver	L	M	Q	H
1	10	8	7	4
2	20	16	12	8
3	32	26	20	15
4	48	38	28	21
5	65	52	37	27
6	82	65	45	36
7	95	75	53	39
8	118	93	66	52
9	141	111	80	60
10	167	131	93	74
11	198	155	109	85
12	226	177	125	96
13	262	204	149	109
14	282	223	159	120
15	320	254	180	136
16	361	277	198	154
17	397	310	224	173
18	442	345	243	191
19	488	384	272	208
20	528	410	297	235
21	572	438	314	248
22	618	480	348	270
23	672	528	376	284
24	721	561	407	315
25	784	614	440	330
26	842	652	462	365
27	902	692	496	385
28	940	732	534	405
29	1002	778	559	430
30	1066	843	604	457
31	1132	894	634	486
32	1201	947	684	518
33	1273	1002	719	553
34	1347	1060	756	590
35	1417	1113	790	605
36	1496	1176	832	647
37	1577	1224	876	673
38	1661	1292	923	701
39	1729	1362	972	750
40	1817	1435	1024	784

Kanji mode: Shift JIS double-byte characters.

Symbol structure

A QR code is not just data modules. Fixed structural regions occupy space in every symbol and are not available for data. Here is what every QR code contains:

Region	Location / size	Purpose
Finder patterns	Three 7×7 squares at top-left, top-right, bottom-left corners	Allow scanners to locate and orient the symbol at any rotation and viewing angle
Separators	1-module white border around each finder pattern	Isolate finder patterns from adjacent data modules
Timing patterns	Alternating black/white row and column connecting the finder patterns	Establish a coordinate grid; help scanners compensate for distortion
Alignment patterns	5×5 squares at calculated positions; version 1 has none, version 40 has 46 centers	Correct for image distortion in larger symbols; count grows with version
Format information	Two copies of 15-bit sequence adjacent to finder patterns	Encodes error correction level and mask pattern; always decodable even when data is damaged
Version information	Two copies of 18-bit sequence (versions 7–40 only)	Encodes the version number
Dark module	One always-dark module at position (row 4 version+9, column 8)	Required by spec; always set to 1
Data + EC modules	All remaining modules	Interleaved data codewords and Reed-Solomon error correction codewords
Quiet zone	Minimum 4-module white margin outside the symbol boundary	Required by ISO/IEC 18004; separates the code from surrounding graphics for reliable detection

How version is chosen

Version is not a user setting -- it is computed by the encoder as the minimum that fits the data. The process:

Identify the most compact encoding mode the input qualifies for (numeric > alphanumeric > byte, or kanji for Shift JIS content).
Compute the total bit count: mode indicator (4 bits) + character count indicator (variable by version and mode) + encoded data bits + terminator bits.
Find the lowest version where available data codeword capacity ≥ required bit count ÷ 8, at the chosen EC level.
Apply one of eight data masks to optimize module distribution, reducing long runs of same-color modules that confuse scanners.

This means shorter data always produces a smaller, simpler QR code. A 20-character URL and a 20-character URL with an extra path segment are not the same size code.

Minimum case: Version 1 at EC L holds 17 bytes in byte mode. A URL like https://example.com (19 bytes) already requires version 2 at EC L, or version 3 at EC M.

Practical limits

The table maximums assume single-mode encoding. In practice, most URLs trigger byte mode (lowercase letters, slashes, query strings). Practical version ranges for common data:

Short URL, ~30 chars: Version 2–3, EC M. Scans from any reasonable distance.
Typical URL, ~60 chars: Version 4–5, EC M. Still compact and reliable.
Long URL, 100+ chars: Version 7+. Consider a URL shortener if printing small.
vCard (business card): Version 6–10 depending on field count.
WiFi credentials: Version 3–5 for typical networks.
Dense text block: Version 20–40. Version 40 EC L holds 2,953 bytes (~2,900 ASCII chars). Very dense; scanning requires a large physical print size.

The real constraint at high versions is physical print size, not the spec limit. Each module in a version 10 code printed at 1 inch (25.4mm) is ~0.44mm. Scanners start struggling below 0.25mm per module. The reliable floor is approximately 1 module = 0.5mm, which sets a minimum print size of about 2.5cm (1 inch) for a version 10 code. Version 40 at that floor requires roughly 9cm (3.5 inches).

For print sizing guidance: QR Code Sizes: What Resolution Do You Need?

Standard reference

All capacity data on this page is from ISO/IEC 18004:2015 — Information technology. Automatic identification and data capture techniques. QR Code bar code symbology specification (second edition). The capacity figures appear in Table 1 (Character capacities of QR Code symbols). Alignment pattern positions are in Table 9.

The standard is maintained by ISO Technical Committee 1, Subcommittee 31 (TC1/SC31). QR code was developed by Denso Wave in 1994; first standardized as ISO/IEC 18004:2000. Denso Wave holds the patent but has chosen not to enforce it, which is why QR codes are freely implementable by anyone.

QR Code Capacity Reference: Versions, Modes, and Character Limits

Versions and module counts

Encoding modes

Capacity tables

Symbol structure

How version is chosen

Practical limits

Standard reference

Continue reading