Matthew 17:21, “this kind goes not out but by prayer and fasting”
Matthew 17:21 is missing from some modern Bibles because they follow the shorter reading ending with the mustard seed; other texts include the sentence about prayer and fasting. The APB retains that sentence in brackets.
The APB text at this verse
The manuscripts, editions, and versions
The committee record at this verse
The APB is translated by a committee of language models that works blind. The translators are never told they are translating the Bible, never given a book name or a verse number, and never shown the witnesses listed above. Where the manuscripts disagree, the choice between readings is made before translation by the same kind of translator working the same way: it sees the competing readings and the witnesses by age and class, but never a reference or a manuscript name. There are 14,877 such places. For APB 4.0 every one was decided again in October 2026. GPT-6.1 Sol decided every one first. Claude Fable 5.1, from another model family, decided 2,483 of them second, including every place where the first translator left the base text. A departure from the base text needs both translators to agree and the first to be at least 0.8 confident. There is no third translator, so the base text stands when the two disagree. APB 4.0 is translated over those decisions. Whatever was done at this verse was done on the language and the evidence alone, and the record is public.
Matthew 17:21: set in brackets as a disputed passage, wording kept.
“The base text reads ὑμῖν. The copies also carry ὑμῖν 21 Τοῦτο δὲ τὸ γένος οὐκ ἐκπορεύεται εἰ μὴ ἐν προσευχῇ καὶ νηστείᾳ. [early greek manuscript tradition; later majority tradition]; the committee kept the base text. The addition is early, but the stronger supplied support and plausible expansion favor the default.”
“The base text reads λέγει. The copies also carry Ἰησοῦς [greek majuscule; later majority tradition]; the committee kept the base text. The earlier support favors retaining the default.”
“The base text reads ὀλιγοπιστίαν. The copies also carry the omission [greek majuscule; later majority tradition]; the committee kept the base text. The default has strong early support and explains the supplied sentence better.”
“The base text reads ἔνθεν. The copies also carry the omission [greek majuscule; later majority tradition]; the committee kept the base text. The strong early support warrants retaining the default.”
“The base text reads λέγει. The copies also carry Ἰησοῦς εἶπεν [greek majuscule; later majority tradition]; the committee kept the base text. The shorter present tense wording has the earlier support. The named subject with a past tense is a copyist's clarification.”
The APB committee decided to set Matthew 17:21 in brackets as a disputed passage and keep its wording. Its note on the added sentence states that “the committee kept the base text.” The two translators were GPT-6.1 Sol (OpenAI), who read first, and Claude Fable 5.1 (Anthropic), who read second. The printed text ends the unbracketed answer with the mustard seed saying, then retains the prayer and fasting sentence in brackets. The passage remains on the page.
The argument for each reading
If it migrated from Mark
Westcott-Hort and NA28 end the answer at Matthew 17:20. In this reading, Jesus answers the disciples with the saying about faith like a grain of mustard. The additional sentence about prayer and fasting does not follow it. The case for this shorter reading rests on that textual support and on a possible explanation for the longer form: a saying from Mark could have entered Matthew through harmonization. Such an expansion would connect the disciples' failure with prayer and fasting as well as faith. This explanation accounts for the added sentence, though it does not by itself establish how every copy acquired it.
If Matthew wrote it
Tregelles retains the sentence about prayer and fasting, and the apparatus places RP alongside it. This gives the longer reading explicit textual support. With the sentence present, Jesus' answer continues beyond the mustard seed saying and identifies prayer and fasting as necessary for this kind to come out. If Matthew wrote it, the shorter reading would leave out that additional instruction. The committee's textual note also acknowledges that the addition is early. Its presence therefore deserves consideration even within an assessment that favors the shorter form. The main uncertainty is whether its transmission preserves Matthew's wording or reflects the movement of a related saying into this account.
The uncertainty concerns whether the prayer and fasting sentence belonged to Matthew's account or entered it through harmonization with Mark. The difference affects the instruction attached to the disciples' failure: the longer reading adds an explicit statement about prayer and fasting. Both readings retain the mustard seed saying about faith. A decision about this sentence alone does not settle a doctrine of prayer or fasting.
Related entries
Witness readings for critical editions are quoted from the SBLGNT apparatus (CC BY 4.0); Hebrew text from the Westminster Leningrad Codex (Open Scriptures Hebrew Bible). Readings of ancient manuscripts and versions are cited from the sources linked above.