For British English speakers, the start of the word new is similar to Spanish ñ. This example doesn't work for most American English speakers, though.
For both British and American English speakers, the middle of the word sinew is fairly similar to the Spanish ñ sound. It's not a very common word, but it is native to English.
Other examples that don't work for "yod-dropping" American English speakers: knew, newt.
/nj/ occurs in many words from Latin-derived languages, including some words that entered English pretty early on
I can't think of any other English words with /nj/ that are not recent and not from Romance languages or Latin. (Well, I guess there may also be some other words that are compounds with /n.j/, like "barnyard": the OED has a quotation from 1473 that uses the spelling "bernȝarde".)
(I interpret "a word that was already present in English dictionaries in the 18th century that didn't come from a Latin-derived language" as excluding both words that are more recent than the 18th century, regardless of their derivation, and words that are from Latin or a Latin-derived language, regardless of their age. If you intended to include Latin-derived words that are older than that, such as the ones mentioned in some of the other answers, you may want to edit the question to make it clearer.)
Phonotactics of /nj/ in yod-dropping accents
One way of interpreting the American English "yod-dropping" change of /nj/ to /n/ in the onset of stressed syllables is as a loss of or a prohibition of tautosyllabic /nj/: if you adopt a certain theory of syllabification, a word like continue can be analyzed as being exempt from yod-dropping because the /n/ is syllabified with the preceding vowel (/kənˈtɪn.ju/), in contrast to a word like continuity, where the /n/ is syllabified with the following vowel. And more controversially, I think, senior and junior could be syllabified as something like /ˈsin.jər/ and /ˈdʒun.jər/ (although I don't really have a strong intuitive sense that this is the correct syllabification of these words—I can only justify it on theoretical grounds). If you adopt such an analysis, there would be no examples in American English of tautosyllabic /nj/ in native vocabulary, so a heterosyllabic sequence /n.j/ would be the closest that you could get.
Despite this, in my experience, American English speakers typically don't have much (if any) trouble producing word-initial (and thus, by necessity, tautosyllabic) /nj/ in foreign words, although some speakers may use a syllabic /ni/ pronunciation instead (possibly with some influence from English spelling conventions where "y" can represent /j/ or /i/, or due to a lesser willingness to use pronunciations that are not fully assimilated to American English phonology/phonotactics). E.g. the American Heritage Dictionary gives the pronunciation of loanword nyala (a type of African antelope) as disyllabic "nyä´lə", while Merriam Webster gives the trisyllabic pronunciation "\ nē-ˈä-lə \".