We　focus　on　the　distributed　statistical　inference　for　linear　models　with　multi-source　massive　heterogeneous　data.　First,　a　communication-efficient　distributed　aggregation　method　is　proposed　to　estimate　the　unknown　parameter　vector,　and　the　derived　estimator　is　proved　to　be　best　linear　unbiased　and　asymptotically　normal　under　some　regularity　conditions.　Then,　a　distributed　test　method　is　proposed　to　test　the　heterogeneity　among　a　large　number　of　data　sources.　Finally,　the　simulations　are　conducted　to　illustrate　the　effectiveness　of　the　proposed　method.　©　2023　Chinese　Academy　of　Sciences.　All　rights　reserved.