Ninety five sans families, and the letterforms are the last thing to choose on

The whole category, on the fields that differ

Weights counts roman styles, with matched italics after the plus. Families with one axis show its range; families with several show the first three axis names and a count of the rest, with full ranges in the profiles below. Rank is the position within the 150 families our census covers. Every name links to its own page with full licence terms.

Nearly half of them have no italic

Forty four of the ninety five families here ship no italic styles. Fifty one ship italics matched to their roman weights.

That is a much higher proportion than in the serif category, and it is not an oversight. A serif italic is a separate historical drawing with different letterforms, so a serif without one is missing something that cannot be faked. A sans italic is usually a slanted and lightly corrected roman, which means the gap between a real one and a synthesised one is smaller, and a designer shipping a Latin family on a deadline can reasonably leave it out.

Smaller is not nothing. Software asked for an italic that does not exist will slant the roman mechanically, which skews the vertical stems and produces the slightly greasy look that makes emphasised text seem cheaper than the text around it. If a design uses italics anywhere, it is worth checking this column before committing, because forty four of these families will silently fake it.

What actually separates a sans here is the script it carries

This page profiles the families the metadata can distinguish, and the result is striking: almost all of them qualify because of script coverage rather than anything about their Latin design.

Twenty six of the ninety five carry a script beyond Latin, Greek and Cyrillic. Broken down, that is seven with Hebrew, five with Arabic, four with Japanese, three with Devanagari, three with Korean, two with Thai, one with Bengali, and one each with Simplified and Traditional Chinese. For comparison, fifty eight carry Vietnamese and thirty seven carry Cyrillic, which are common enough to distinguish nobody.

The practical reading is that if a project is Latin-only, the choice between most of these families is genuinely a matter of taste and the data will not help. The moment a project needs Arabic, Hebrew, Devanagari, Japanese or Korean, the field of ninety five collapses to a handful, and the choice makes itself. That is the single most useful thing this category has to tell you, and it is invisible until the coverage is counted.

One family exposes thirteen axes

The category norm is a single weight axis. Roboto Flex exposes thirteen: weight, width, optical size, slant, grade, and eight more that almost nothing else in the corpus offers.

Those eight are parametric axes, and they are a genuinely different idea about what a typeface is. Rather than exposing finished design decisions such as bold or condensed, they expose the underlying parameters a designer would have manipulated to reach them: the thickness of vertical strokes and of horizontal strokes as separate quantities, the counter width, and the heights of ascenders, descenders, lowercase, uppercase and figures each independently adjustable.

The honest assessment is that this is more control than almost any project needs and more than most tools expose. It matters here for a narrower reason: it means a single family can be tuned to sit correctly next to something else, matching an existing x-height or figure height rather than approximating it. That is a real problem in title design and it usually gets solved by choosing a different typeface.

Two rows in this category are not sans serif faces

Noto Color Emoji is filed here. It is an emoji font, it covers the emoji subset and nothing else, and at roughly 23,700KB it is the largest file in the category by a factor of twelve against the next largest. It has one style and no italic. Nothing about it is a sans serif.

Noto Sans Mono is also filed here, and it is a monospace family, which is a separate category in this same corpus containing seven other families. It is a capable face with nine weights, a width axis and coverage across Greek and Cyrillic. It simply is not a sans serif in the sense the other ninety three rows mean.

Both are left in the table rather than quietly removed. The category is what the source says it is, and a roundup that silently edits its own source is worth less than one that shows the outliers and explains them. It is also a reminder worth carrying: a category label in a dataset is a claim by whoever assigned it, not a fact about the object.

The width axis, and the fourteen that have it

Fourteen of the ninety five expose a width axis, and five expose an optical size axis.

The width axis is the one that solves a concrete layout problem: a headline that overruns its column can be condensed rather than shrunk, and it stays the same typeface at the same weight. Doing that without an axis means either dropping the point size, which weakens the headline everywhere, or switching to a condensed sibling, which is a different drawing and rarely matches exactly.

It is worth being clear about what a real width axis is not. Squeezing type horizontally in a design tool scales the vertical strokes along with the horizontal ones and produces the thin, smeared verticals that mark out amateur work at a glance. A drawn width axis holds stem weight across its range because the designer decided what the letters should look like at every width. The axis is the difference between condensing type and pretending to.

File size spans a factor of nearly five hundred

The smallest family here is 50KB and the largest is roughly 23,700KB, against a category median of 194KB.

As in the serif category, the extremes are doing a different job. Noto Color Emoji carries thousands of multicolour glyphs. The CJK families, LINE Seed JP, Gothic A1 and BIZ UDPGothic, carry the thousands of characters their scripts require. None of them is bloated and none of them is comparable with a 194KB Latin family.

The number worth holding is the median. If a Latin-only sans in this corpus is much heavier than about 200KB, it is usually carrying either extended script coverage or several variable axes, and both are things you can check for rather than guess at.

The 12 the data actually distinguishes

The table above holds all 95 families in the category, and every one of them links to its own page. The 12 below are the ones carrying at least two measured facts that are true of them and false of most of their siblings. The remainder are not worse typefaces. They are families the metadata cannot separate, and inventing a distinction for each of them would be the opposite of useful.

Common questions

How many sans serif fonts are in the 150 most-used Google Fonts?

Ninety five, which makes Sans Serif much the largest category in that set. Serif is second at twenty six, then Display at thirteen, Handwriting at eight and Monospace at seven. Open Sans and Roboto are tied as the highest-ranked families in the whole corpus.

How many of them have no italic?

Forty four of the ninety five ship no italic styles at all, and fifty one ship italics matched to their roman weights. Software asked for an italic that does not exist will slant the roman mechanically, which distorts the vertical strokes. If a design uses italics anywhere, this is worth checking before committing to a family.

Which sans serif fonts cover Arabic, Hebrew or CJK scripts?

Twenty six of the ninety five carry a script beyond Latin, Greek and Cyrillic: seven with Hebrew, five with Arabic, four with Japanese, three with Devanagari, three with Korean, two with Thai, one with Bengali, and one each with Simplified and Traditional Chinese. Rubik is the only family carrying Arabic and Hebrew together. That coverage is the field that most usefully narrows a choice among ninety five otherwise close options.

Which font has the most variable axes?

Roboto Flex, with thirteen, against a category norm of one. Alongside weight, width, optical size, slant and grade it exposes eight parametric axes covering stroke thickness, counter width and the heights of ascenders, descenders, lowercase, uppercase and figures. It is by a wide margin the most adjustable family in the corpus.

Why is Noto Color Emoji listed as a sans serif?

Because the source metadata files it that way. It is an emoji font covering only the emoji subset, with one style and no italic, and at roughly 23,700KB it is twelve times larger than the next biggest file in the category. Noto Sans Mono is filed here too and is a monospace family. Both are left in the table rather than removed, because a category label in a dataset is a claim by whoever assigned it and not a fact about the object.

Are these sans serif fonts free for commercial use?

Every family in the 150 most-used set is under an open licence, almost all of them the SIL Open Font License 1.1, which permits commercial use, monetised video, broadcast, merchandise, client work and logos with no attribution required. The licence forbids selling the font software on its own and requires a modified version to be renamed. Each family page on this site records its exact licence with a source URL and a verification date.

Method and sources. Category, licence, designer credits and popularity rank come from our own typeface census, which reads the Google Fonts metadata API and the licence files in the google/fonts repository, and records a licence only with a source URL and a verification date. Weights, italics, variable axes, script subsets and byte size come from the Google Fonts metadata API read on the build date shown above. Popularity rank is the position within the 150 families the census covers, not the global ordering. With ninety five families in this category, individual profiles are reserved for the twelve carrying at least four measured facts that are true of them and false of most of their siblings; the full category is in the table and every family in it links to its own page. That threshold is a statement about what the metadata can distinguish, not a ranking of the typefaces. Built 2026-08-29.