Unreleased stops

/t/ is not released when syllable-final
/p/ and /k/ are released when syllable-final

David Kahn in his "Syllabic generalizations in AmE Phonology" notes it. Even observing native speakers attest this fact. Chinese speakers unrelease /p/ and /k/ when syllable-final: you can contrast it with the native speaker's

On side note, Voice Onset Time (VOT) for /k/ is longer than others. cf:

John Maidment says
"One might also like to add that VOT is sensitive to place of articulation. Other things being equal, the VOT of posterior articulations, velar and uvular, are considerably longer, at least for native English speakers, than articulations further forward in the vocal tract. A typical VOT for stressed syllable initial [k] in English is in the region of 120ms, while that for an equivalent [p]is only 60-70ms. I am pretty sure that this difference is an important secondary cue for the perception of place in voiceless plosives and one which, as far as I know, has not been properly investigated.

The reason for this difference has a plausible aerodynamic explanation, but I perhaps should not go into this now."