SMS character limit: why one text is billed as 2 or 3

An SMS holds 160 characters only while every character in it belongs to the GSM-7 alphabet. One character outside that alphabet, such as a Polish ł, a Spanish á, an emoji or a curly apostrophe, switches the whole message to UCS-2 encoding, which fits just 70 characters, and long messages then split every 67 characters instead of every 153. Providers bill per segment, so that single character can double or triple the cost of a send.

Why is an SMS 160 characters, or only 70?

Because one SMS segment has a fixed payload of 140 bytes (1,120 bits), and the encoding decides how many characters fit into it. GSM-7 uses 7 bits per character, so 1,120 ÷ 7 = 160 characters; UCS-2 uses 16 bits per character, so 1,120 ÷ 16 = 70 characters.

The AWS End User Messaging SMS documentation puts it plainly: a message that uses only characters from the GSM 03.38 set can contain up to 160 characters, and a message with any character outside that set can have up to 70. The switch is all-or-nothing. The encoding is chosen per message, not per character, so 159 plain letters and one emoji are encoded entirely in UCS-2.

Why do long messages split at 153 and 67 characters?

When a message is longer than one segment, each segment carries a User Data Header (UDH) that tells the phone how to reassemble the parts, and that header takes space from the text. In GSM-7 it costs 7 characters, which leaves 153 per segment; in UCS-2 it costs 3 characters, which leaves 67.

Twilio's SMS character limit page gives the same figures: 160 for a single GSM-7 segment, 153 per segment in a multi-part message, 70 for a single UCS-2 segment and 67 per segment after that. It also notes an exception for US and Canadian toll-free numbers, where the limits are 152 and 66.

The practical consequence is that a 161-character message is not "one SMS plus one character". It becomes two segments of up to 153 characters, and the capacity of two segments is 306 characters, not 320. The table shows where each extra segment starts.

Calculated from 160/153 and 70/67 characters per segment. As of November 2026.

Providers also cap the total length. SMSAPI accepts up to 1,530 characters (670 with special characters) in at most 10 parts, AWS accepts 1,530 GSM characters or 630 non-GSM characters, and Twilio lists 1,600 and 700, while recommending that you keep SMS under 320 characters. Longer messages are rejected or, depending on the provider, sent another way, so check your provider's limit before you build long templates.

Which characters force UCS-2 encoding?

Any character that is not in the GSM 03.38 table. For most senders the culprits fall into three groups: letters with diacritics that GSM-7 does not include, typographic punctuation inserted by word processors, and emoji.

The AWS page reproduces the full GSM 03.38 table. Comparing it with the alphabets of a few European languages shows why the same campaign can cost one segment in German and two in Polish:

Based on the GSM 03.38 character table published by AWS. As of November 2026.

Characters that slip in without anyone typing them deliberately cause most of the surprises:

• Curly quotes and apostrophes (’ ‘ “ ”): Word, Google Docs and macOS replace straight quotes automatically. The AWS documentation shows a 130-character message that needs two parts only because of one closing single quote in "Dr. Salazar’s".
• Dashes and ellipsis (– — …): the hyphen-minus (-) is in GSM-7, the en dash and the single-character ellipsis are not.
• Invisible characters: a non-breaking space or a zero-width space copied from a web page or a CMS looks like a normal space but forces UCS-2.
• Emoji: no emoji is in GSM-7. According to Twilio's emoji support article, an emoji uses 2 of the 70 characters of a UCS-2 segment, and emoji with modifiers take even more.

⚠️ Watch out: Personalisation changes the encoding per recipient. "Hi [[fname]]" stays in GSM-7 for Anna and switches to UCS-2 for Łukasz or Ramón, so the same template can cost one segment for one contact and two for another.

Which GSM-7 characters count as two?

Nine printable characters live in the GSM-7 extension table and each is sent as an escape character plus the character itself, so it takes two of the 160 (or 153) places: ^ { } \ [ ] ~ | €.

These characters do not switch the message to UCS-2, but they do eat into the limit. "Price: 20€ [today only]" is 23 characters on screen and 26 in GSM-7. The effect is small in plain text and large in messages that contain JSON, code snippets or URLs with brackets. If a message hovers near a boundary, write "EUR" instead of "€" or replace square brackets with round ones.

If a counter and the table ever disagree, trust the segment count your provider returns after sending, because that is what you are billed for.

How much does the encoding add to the bill?

Every segment is billed as a message. Twilio states that it "bills SMS per message segment sent", AWS charges for the number of message parts, and SMSAPI sends long messages as up to 10 linked parts billed according to its price list. A message that grows from one to two segments doubles the cost of the send.

The table multiplies official per-segment prices by 1,000 recipients. The prices are list prices for outbound SMS on the providers' public pages; volume discounts, carrier fees and VAT are not included.

Twilio prices "current as of October 2026" on Twilio's pages; SMSAPI price list checked November 2026. As of November 2026.

The example below shows how little it takes. Both texts say the same thing in Polish; the first is 94 characters with diacritics and goes out as two UCS-2 segments, the second is the same 94 characters without them and fits into one GSM-7 segment:

Cześć Anna, Twoje zamówienie 1042 zostało wysłane. Śledź: https://ex.co/t/1042 STOP=rezygnacja
→ UCS-2, 94 characters, 2 segments

Czesc Anna, Twoje zamowienie 1042 zostalo wyslane. Sledz: https://ex.co/t/1042 STOP=rezygnacja
→ GSM-7, 94 characters, 1 segment

The German equivalent, "Hallo Jürgen, deine Bestellung 1042 ist unterwegs…", stays in GSM-7 at 109 characters, because ü is in the table. The Spanish one with "está" and "Síguelo" switches to UCS-2 at 106 characters and needs two segments.

Should you strip diacritics from SMS text?

It depends on the language and on the reader. Stripping diacritics (transliteration) keeps a message in GSM-7 and can halve the cost, but in some languages it lowers readability and can look careless, so it is a trade-off, not a rule.

You do not need to strip diacritics by hand. SMSAPI offers a normalize parameter that replaces special characters with their plain equivalents (ą→a, ć→c, ę→e), and its documentation says the default encoding is windows-1250 unless you pass another one such as utf-8. Twilio offers Smart Encoding, which replaces some non-GSM characters with GSM equivalents and turns curly quotes into straight ones.

A pragmatic middle ground many teams use: keep diacritics in short messages that fit into 70 characters anyway, and transliterate only when the text would otherwise need an extra segment. Whatever you choose, decide it once per language and apply it consistently.

Do links, opt-out text and the sender ID count?

Links and opt-out text count toward the length because they are part of the message body; the alphanumeric sender ID does not, because it travels in the address field of the SMS, not in the text.

Links

A full URL with UTM parameters easily takes 80–120 characters, more than half a GSM-7 segment. Use a short path on your own domain (for example https://ex.co/t/1042) rather than a long tracking URL. A short link on your own domain also keeps your brand visible and does not depend on a third-party shortener.

Opt-out text

Marketing SMS usually carry an opt-out instruction; the exact rules depend on the country. "Reply STOP to opt out" is 21 characters; budget for it in every template rather than adding it at the end and discovering a second segment. Keep it in GSM-7 characters so it never triggers UCS-2 by itself. For the consent rules that apply in your country, see the guide to email and SMS automation on your own server. This is not legal advice.

Placeholders

Placeholders such as a first name, an order number or a coupon code make the final length variable. Test templates with your longest realistic values, not with "Anna", and remember that a placeholder value with a diacritic changes the encoding for that recipient.

How do you check the segment count before sending?

Run the final text, with placeholders filled in, through a counter that knows the GSM 03.38 table, then confirm with your provider on a test send. The provider's figure is the one you pay for.

1. Find non-GSM characters

   Paste the text into a counter that lists the characters that force UCS-2, not only the total. The Python function below does that offline.

2. Test with the longest real values

   Fill placeholders with the longest names and codes from your data, and with names that contain diacritics.

3. Send a test SMS to your own phone

   Check how it renders, whether the parts arrive joined and whether the link works.

4. Read the segment count from the provider

   Twilio returns num_segments in the message resource; SMSAPI returns parts and the cost in points. Compare it with your estimate before sending to the whole list.

This function follows the GSM 03.38 table from the AWS documentation. It counts extension characters as two and emoji as two UCS-2 units. It is an estimate: real gateways may not split an escaped character or an emoji across two segments, which can occasionally add a segment near a boundary.

import math

GSM7_BASIC = set(
    "@£$¥èéùìòÇ\nØø\rÅåΔ_ΦΓΛΩΠΨΣΘΞÆæßÉ !\"#¤%&'()*+,-./0123456789:;<=>?"
    "¡ABCDEFGHIJKLMNOPQRSTUVWXYZÄÖÑܧ¿abcdefghijklmnopqrstuvwxyzäöñüà"
)
GSM7_EXTENDED = set("\f^{}\\[]~|€")  # each costs 2 septets (ESC + char)


def sms_segments(text: str) -> dict:
    non_gsm = sorted({c for c in text if c not in GSM7_BASIC | GSM7_EXTENDED})
    if not non_gsm:
        units = sum(2 if c in GSM7_EXTENDED else 1 for c in text)
        single, per_part, encoding = 160, 153, "GSM-7"
    else:
        units = len(text.encode("utf-16-le")) // 2  # emoji = 2 units
        single, per_part, encoding = 70, 67, "UCS-2"
    parts = 1 if units <= single else math.ceil(units / per_part)
    return {"encoding": encoding, "units": units, "parts": parts,
            "non_gsm": non_gsm}


if __name__ == "__main__":
    import sys
    print(sms_segments(sys.stdin.read().rstrip("\n")))

Save it as sms_segments.py and pipe a message into it:

echo "Don’t miss it" | python3 sms_segments.py
# {'encoding': 'UCS-2', 'units': 13, 'parts': 1, 'non_gsm': ['’']}

The non_gsm list tells you exactly which character to replace. In the example above, swapping the curly apostrophe for a straight one brings the message back to GSM-7.

Segment counts when you send SMS from NetSendo

NetSendo sends SMS through Twilio, Vonage, MessageBird, Plivo or SMSAPI, and the segment rules above are the provider's, not NetSendo's. Two details matter in practice.

First, the counter in the SMS editor is a quick estimate. It labels a message as Unicode as soon as it contains any non-ASCII character and divides the length by 160 or 70. That means a German message with ü is shown as Unicode, although GSM-7 includes ü, and messages just above a segment boundary can be counted one segment short because the 153/67 rule is not applied. Use it as a warning light and check borderline texts with the function above.

Second, the provider's real figure is available after sending. The sms.sent webhook includes parts and credits for each message, taken from the provider's response (SMSAPI's parts and points, Twilio's num_segments). You can collect it in n8n with the NetSendo node or any endpoint described in the webhooks documentation, verify the HMAC-SHA256 signature and alert when a campaign averages more segments than planned. NetSendo sends to SMSAPI with encoding=utf-8, so diacritics are delivered as written; if you want them stripped, do it in the text. Messages can also be sent through the REST API.

ℹ️ Note: SMS sending in NetSendo was broken in older releases; version 2.2.0 fixed the SMS job and 2.7.2 fixed campaigns sent from the panel and test SMS. Update to the current release before you rely on SMS.

Frequently asked questions

Sources

• Twilio: What is the SMS character limit?
• Twilio Support: Can I send and receive emoji with Twilio?
• AWS End User Messaging SMS: SMS character limits and GSM 03.38 character set
• SMSAPI documentation (message, encoding, normalize, max_parts)
• SMSAPI price list
• Twilio SMS pricing: Poland
• Twilio SMS pricing: Germany
• Twilio SMS pricing: Spain