This study investigated the effects of hybrid feedback — a combination of recast and metalinguistic clues — on the grammatical accuracy of intermediate L2 learners' speech production, relative to either feedback type alone. Forty-five Iranian EFL learners (B1/intermediate level) were randomly assigned to a recast group, a metalinguistic-clues group, or a hybrid group and received one of the three feedback types over nine weekly speaking sessions targeting countable and uncountable nouns. Implicit and explicit grammatical knowledge were assessed before and after the intervention using an elicitation-imitation (EI) task and a grammaticality judgment test (GJT), respectively. A two-way mixed ANOVA revealed significant Time × Group interactions for both measures, and follow-up one-way ANOVAs showed significant group differences at posttest on both EI, F(2, 42) = 4.31, p = .020, and GJT, F(2, 42) = 4.99, p = .011. Post hoc comparisons indicated that the hybrid group significantly outperformed the recast group on both measures, while the metalinguistic group did not differ significantly from either. All three groups improved over time, but gains were most pronounced for hybrid feedback, followed by metalinguistic clues, with recast producing the smallest gains. These findings suggest that combining implicit and explicit correction can be more effective than either form in isolation, particularly for grammatically fossilized structures such as countable and uncountable nouns, and carry implications for how oral corrective feedback is delivered in L2 classrooms.