Reading levels by grade: Peabody's 13 levels on one page

Peabody (peabody.kids) sorts every story into one of 13 reading levels, plus a level 0 floor for the very first books. A level is measured from the text, never from the child: the longest sentence, the words on a screen, a Flesch–Kincaid range for the paragraph levels, a ceiling on the share of words off the Dale–Chall familiar-word list, and for the early levels the sight words and the phonics patterns a beginner can decode. Level 1 is "I see a boy." and level 10 is about an eighth-grade paragraph. Peabody is a free AI-driven adaptive learning platform for kids, free through the end of 2026, and this cheat sheet is the whole level table with the limits the code enforces. AI writes the stories; code decides the rest.

How to use this: find the row that matches the longest sentence your child reads without stopping, then read the "how to pick a starting level" section before you trust the grade column. The grade column is the roughest number on the page.

The 13 levels (and the floor)

The limits below are the ones the code checks. "Longest sentence" is a hard limit on any sentence in a story at that level. "Words per screen" is the range a screen is written to. The Flesch–Kincaid (FK) column is the range the whole story must land in; the early levels have none, because the formula does not work on three-word sentences. The Dale–Chall (DC) column is the most words a story may have off the familiar-word list. The example sentences are mine, written in each level's style.

Three details that matter when you read the table:

• Level 0 is a library floor, not a level a child is placed at. The free library has first books there (one or two sentences a page, one to four words each) for a child who is not yet reading level 1. The story writer never writes at level 0, and placement never starts a reader there.
• Levels 1 and 2 also count sentences per screen: one or two at level 1, two or three at level 2. Level 2 is the bridge from a line to a paragraph.
• The writer aims under each level's middle. A language model writes harder than it is asked, so the story writer is given the easy half of the FK range and a short average sentence (about 3 words at level 1, 6 at level 2, 10 at level 3, rising to 20 at level 13). The level's limits are the gate: a story over its ceiling is repaired or rewritten, and one under it gets a warning.

The library's 201 hand-written stories, 110 of them retold classics, are measured the same way and sit at every level from 0 to 8th grade. All but two of the scarier ones are open without an account in the free library.

What each measure actually computes

Flesch–Kincaid grade level is a weighted sum of two averages: words per sentence and syllables per word. Nothing else goes in. The output is scaled so that it looks like a US school grade, which is convenient and misleading in equal measure: a text of short sentences made of rare one-syllable words scores as easy, and a text of long sentences made of familiar words scores as hard. Day and colleagues (2025) found that language models asked to simplify text to a grade 4 target by this formula missed it in every model and prompt combination they tried, partly because the formula rewards short sentences whatever the words in them (Readability Consortium preprint). That is why Peabody uses FK as a range and never alone.

Dale–Chall replaces syllables with a list. It counts the share of words that are not on a list of about 3,000 words that most fourth graders know, and combines that with sentence length. Peabody uses the share on its own as a ceiling per level, from 4% at level 0 to 24% at level 13, so that a story cannot pass on sentence length while being full of words a child has never met. The seed library sits well under its ceilings.

Why any formula wanders about a grade on a short text. A formula is an average, and an average of eighty words is noisy. One long sentence of dialogue, one three-syllable name, or one list moves a short text's FK by a grade in either direction. Peabody's stories are short, so every story gets a range rather than a point, the FK ranges of neighbouring levels overlap at the bottom (level 2 runs 1.0–3.0 and level 3 runs 2.5–3.5), and a library label is picked from the FK score by a fixed boundary table. Lexile, the other common text measure, works the same way: sentence length plus word frequency, as MetaMetrics describes it (Lexile measures overview). Peabody doesn't measure Lexile.

Why the early levels use sentence length and decodability instead. Below about level 3 the formulas have nothing to measure: every word is one syllable, every sentence is a few words, and the score comes out negative. What matters for a beginner is whether the words can be sounded out with the letter patterns they have been taught, and how many of the rest are sight words. So levels 0 to 2 are set by the longest sentence, the words on a screen, repetition, sight words, and decodability at the child's phonics step (Peabody has a 16-step phonics sequence, from s a t p i n to -ing and -ed). Cheatham and Allor (2012, Reading and Writing) reviewed the decodable-text studies and found that decodability raises the use of sounding-out strategies and accuracy in beginners, while arguing that it should be one criterion among several rather than the only one (doi:10.1007/s11145-011-9355-2). That is the stance the early levels take. The research report behind this guide also notes that nothing has been published on controlling language-model text at the decodable level, which is why the early levels are checked by pattern, not by formula.

How other systems label levels

Schools and publishers use several scales. The table below relates them to Peabody levels by grade, and only by grade, because that is the one thing the scales have in common. Every column is approximate; the publishers' own grade charts disagree with each other, and a child's placement on any of them comes from a different kind of test. Peabody doesn't measure Lexile, doesn't assign a guided reading letter or a DRA number, and never reports a reading age or a grade equivalent.

What each scale is measuring:

• Lexile gives a text a number from sentence length and word frequency, and gives a reader a number from a comprehension test; a match is meant to produce about 75% comprehension. It says nothing about content, decodability or whether the child can sound out the words. It is a trademarked scale, so a product cannot give out Lexiles without a licence, and Peabody doesn't measure Lexile.
• Guided reading levels A–Z are assigned by a publisher's judgement of text features (layout, vocabulary, sentence complexity, picture support) and a child's level comes from a running record, where a teacher listens to the child read and counts errors. A 2006 generalisability study found a child must read at least three passages for a reliable running-record score (Utah State University).
• DRA (Developmental Reading Assessment) is a similar one-to-one reading conference scored for accuracy, fluency and comprehension, with numbers instead of letters.
• Oral reading fluency screens such as DIBELS count words read correctly per minute against grade benchmarks (DIBELS 8 technical manual). Peabody cannot do this at all: it never records a child, so it cannot hear reading or measure fluency.

The honest summary is that the scales agree on the shape (text gets longer, sentences get longer, words get rarer) and disagree on everything else. Castles, Rastle and Nation (2018, Psychological Science in the Public Interest) put the deeper reason plainly: decoding and language comprehension are separate skills, and no text measure captures the second (Ending the reading wars).

How to pick a starting level at home

Peabody has two ways in, and neither is a test.

1. The sentence tap. A child is shown a short ladder of sentences, one per level, and taps the one they can read comfortably. That sets the starting level. If your child is next to you, ask them to read the sentence out loud before tapping; the one they read without slowing is the one to tap, not the hardest one they can get through.
2. The five-tap check. "Let's find your books" is a five-question game with no score and nothing announced. It gives the code a first guess, and the first few stories refine it.

After that the level moves with the books. If recent books have all run too easy or all too hard, Peabody offers a re-check next door; nothing is graded, and a child is never told they went down. In a free grown-up account you can see which books ran too easy or too hard, in plain words.

Two home checks that are worth more than any formula:

• Listen to a page. Pick a story one level up from where your child is placed and have them read a screen to you. More than about one stumble in ten words means the level below is the one to read for a while; none at all for a whole book means try the one above.
• Ask one question after. Peabody's stories end with four to six comprehension questions, read aloud on request. If your child decodes the words but cannot say what happened, the problem is language, not level, and the fix is reading to them and talking, not an easier book.

The research report behind this guide makes the same point about placement from the testing side: a five-question check cannot narrow a child's position to better than about half a logit even with ideal questions, so a short placement is a prior, not a diagnosis, and practice is what refines it. Sénéchal and Young (2008, Review of Educational Research) found that parents listening to their child read helped more than parents reading to the child, and parents teaching a specific skill helped most of all (meta-analysis summary). The level table tells you which book; listening tells you whether it was right.

What this table doesn't tell you

• It does not measure your child. Every number above is a property of a text. The child's level is a pointer to a row, moved by which books ran easy or hard.
• It does not know what a word means to your child. A word on the familiar list can still be new, and a word off it (dinosaur, volcano) can be the one a child knows best.
• It cannot hear reading. Peabody does not record a child, so accuracy, speed and expression are invisible to it. If you want words-per-minute, a teacher with a stopwatch has it and Peabody does not.
• It does not diagnose. Reading difficulties and dyslexia are found by a specialist with the right tests, never by a screen-tap check.
• The grade column is a convention. FK is scaled to look like a grade, and publishers' grade bands disagree. Treat "2nd" as "the kind of paragraph many second graders read", not as a judgement.
• Peabody has no published efficacy study. The levels are a careful way to label text; whether reading at them helps a particular child is not something Peabody has tested.

For what the level number means in practice, read what a reading level means, and for the rules that move a child between levels, how Peabody picks the next question. The quickest check of all is to open the free library and let your child tap a book.

Each level has its own page with a real sample story, the skills practised there and the free games: start at reading levels and math by grade, or open level 1, level 4 or level 8.

Questions parents ask

My child is in 2nd grade but placed at level 2. Is that bad?

No. Level 2 is where most children read in the first half of first grade, and a child a year behind the grade column is reading the kind of sentence that many classmates also read. The placement is a pointer that moves with the books; if the first few stories run easy, the level will climb. The number to watch is not the level but whether your child wants the next story.

Why do the Flesch–Kincaid numbers look higher than the grades?

Because FK is a formula scaled to a grade long ago on adult texts, and short children's texts score about one grade high on it. Peabody's level 3 runs FK 2.5–3.5 and is called first grade because of what first graders in the hand-written library actually read, not because of the formula. The grade column is the rough match; the FK column is the limit the code checks.

Can I get my child's Lexile from Peabody?

No. Peabody doesn't measure Lexile, and it does not report a grade equivalent or a reading age either. It reports what it measured: which books ran too easy or too hard, which practice words are known, learning or missed, and how the comprehension questions went.

Does the AI decide the level?

No. AI writes the story; code measures it against the table above and repairs or rewrites a story over its ceiling. A child's level is never a language model's opinion.

What if the sentence tap is wrong?

Then the first stories will run too easy or too hard, and Peabody offers a re-check next door. A grown-up can also pick a different starting sentence at any time. Nothing is locked.