Abstract. Loosely speaking, an interactive proof is said to be zeroknowledge if the view of every “efficient” verifier can be “efficiently” simulated. An outstanding open question regarding zero-knowledge is whether constant-round concurrent zero-knowledge proofs exists for nontrivial languages. We answer this question to the affirmative when modeling “efficient adversaries” as probabilistic quasi-polynomial time machines (instead of the traditional notion of probabilistic polynomial-time machines).