[Agent Tools](https://tools.openkrill.app/) / [Blog](https://tools.openkrill.app/blog)

# How many SMS segments is a 161 character text? Two, not one

How many SMS segments is a 161 character text? Two, in our run. A 160-character booking reminder was 1 segment, one more character made it 2, and one emoji made it 3 even though the character count stayed at 160. If you pay per segment, character count alone will mislead you.

By OpenKrill. Published 2026-11-12. Based on our own checks.

## Method and exact input

This is our own experiment, run on 2026-11-12 (GMT+7). We called the count_text tool as a normal caller, with public inputs only and no key, in 8 calls on one invented reminder and its variants, changing one thing per variant against call 1. Call 1 was this text, 160 characters:

```
{"text":"Hi Sam, your table for four is booked for Friday at 7 pm. Reply YES to confirm, or call 555 0100 to change the time. Bring the booking code 4821. See you at Lui"}
```

It returned 160 characters, 160 UTF-8 bytes, GSM-7, 160 units, 1 segment.

## What each variant returned

| Call | Change | Encoding | Segments |
| --- | --- | --- | --- |
| 1 | none | GSM-7, 160 units | 1 |
| 2 | final "!" added, 161 characters | GSM-7, 161 units | 2 |
| 3 | last character swapped for 1 emoji (Lu😀) | UCS-2, 161 units | 3 |
| 7 | Sam swapped for Zoé | GSM-7 | 1 |
| 8 | YES put inside 2 curly quotes | UCS-2 | 3 |

Call 3 is the surprising one. It still has 160 characters and 160 graphemes, but 163 UTF-8 bytes. The emoji forced UCS-2, where one segment holds 70 units (67 when split) instead of 160 (153 when split). Call 8 shows the same jump from curly quotes, which are easy to introduce by pasting from a word processor. The accented name in call 7 did not change anything: é is in the GSM-7 set.

## What to do with a template

1. Write the reminder with its longest realistic values.
2. Run it through the counter and note the encoding, units and segments.
3. Run each variant you can predict: names with accents, an emoji, curly quotes.
4. If a variant flips the encoding to UCS-2, or the units pass 160 in GSM-7, shorten the template or strip that character before sending.

In our run the fix is simple: the plain text sits exactly at the 160-unit edge, so any added character costs a second segment. Leaving a few units of headroom, or replacing the emoji and curly quotes with plain ones, keeps the text at 1 segment.

You can paste your own reminder into [Document Tools](https://tools.openkrill.app/tools/docs), which reports the encoding, units and segments for free with no key, so you see the effect before you pay for it.

## Where this stops

The counter applies the rules it states: GSM-7 holds 160 units (153 when split), UCS-2 holds 70 (67 when split), and characters from the GSM extension table count 2 units. We did not send any message, so we did not check what a carrier or provider actually bills. We ran no rival counter, so we claim no comparison.

Two honest losses. Calls 4 and 5 repeated the input of call 1 by our own mistake, because the swap we meant to make did not match the text, so they add nothing. Calls 7 and 8 are the corrected inputs. Also, the response includes a reading ease score of 100.5 from a syllable estimate, which is meaningful for English only and irrelevant to SMS cost.

This helps anyone sending templated texts per segment. It does not help if your provider bills a flat rate per message.

## Frequently asked questions

### What does one call return for a reminder text?

One call returned characters, graphemes, UTF-8 bytes, words and an sms block with the encoding, units and segments. For call 1 that was 160 characters, 160 bytes, GSM-7, 160 units and 1 segment.

### Is an emoji counted as one character?

Yes for characters and graphemes: call 3 showed 160 of each. But it was 163 UTF-8 bytes and the text counted 161 units in UCS-2, which gave 3 segments.

### Is there a size limit on the text?

Yes. A 20,005-character text (call 6, the word "word" and a space 4,001 times) was refused with invalid_input, because the tool takes at most 20,000 characters.

## Sources

- Based on our own checks.
