Blasting　is　the　primary　technical　method　of　mining　mineral　resources,　the　accurate　prediction　of　blasting　fragmentation　after　blasting　has　a　key　impact　on　controlling　mining　operation　costs　and　reducing　environmental　pollution.　To　that　end,　a　powerful　artificial　intelligence　(AI)　model,　named　Gaussian　Process　Regression　(GPR),　was　proposed　to　predict　the　mean　fragment　size　(MFS)　of　broken　rock　after　blasting.　To　improve　the　prediction　performance　of　GPR　model,　the　newly　bio-inspired　Artificial　Hummingbirds　Algorithm　(AHA)　was　employed　to　select　the　optimal　hyperparameter’s　combinations.　Additionally,　the　other　regression　models　including　the　artificial　neural　networks　(ANN),　support　vector　regression　(SVR),　and　standard　GPR,　were　also　developed　to　compare　the　prediction　performance　with　the　proposed　hybrid　AHA-GPR　model.　Two　cases　collected　from　Chilean　copper　mine　and　an　Indian　coal　mine　were　used　to　verify　the　model　performance,　mainly　using　two　performance　metrics　and　the　Taylor　diagram　analysis　to　determine　the　best　prediction　model.　The　AHA-GPR　model　achieved　R²　values　of　0.88　and　0.85,　with　RMSE　values　of　11.40　and　9.50　for　the　Copper　mine　database,　and　R²　values　of　0.93　and　0.90,　with　RMSE　values　of　0.02　and　0.03　for　the　Coal　mine　database.　These　results　demonstrate　that　the　AHA-GPR　model　outperforms　other　predictive　models,　establishing　it　as　the　most　reliable　approach　in　both　case　studies.　Besides,　this　AI　model-based　approach　offers　researchers　and　mining　engineers　an　improved　approach　to　predict　the　rock　fragmentation　after　blasting,　facilitating　thorough　analysis　of　blasting　design　before　mining　operations　and　enabling　precise　forecasting　of　blasting　outcomes.　Furthermore,　the　sensitivity　analysis　further　provided　essential　insights　to　optimize　the　blasting　designs.　©　The　Author(s),　under　exclusive　licence　to　Springer-Verlag　GmbH　Germany,　part　of　Springer　Nature　2025.