Scientific　evaluation　for　the　efficiency　of　green　low-carbon　innovative　development　(GLID)　has　great　significance　for　promoting　sustainable　development　(SD).　The　connotation　and　composition　of　GLID　were　first　analyzed,　and　GLID　was　regarded　as　a　two-stage　complex　system,　in　which　multi-input　coexists　with　multi-output.　Based　on　the　existing　methods,　an　improved　two-stage　non-cooperative　data　envelopment　analysis　(ITNDEA)　is　further　proposed　to　measure　GLID　efficiency　by　taking　30　Chinese　provinces　in　2020　as　cases,　and　the　results　are　finally　compared　with　that　by　two-stage　data　envelopment　analysis　(TDEA)　and　two-stage　non-cooperative　data　envelopment　analysis　(TNDEA).　The　results　show　that:　(1)　ITNDEA　effectively　avoids　the　efficiency　being　overestimated　or　underestimated　and　can　obtain　a　more　objective　and　reasonable　efficiency　value　than　TDEA　and　TNDEA;　(2)　The　main　reason　of　the　low　GLID　efficiency　is　the　imbalance　between　green　technology　innovation　and　carbon　emission;　(3)　Each　policy　goal　has　a　good　effect　on　efficiency　improvement,　but　the　effect　of　the　same　policy　goal　has　regional　difference;　(4)　All　regions/countries　can　improve　SD　level　through　key　breakthrough　mode　and　gradual　breakthrough　mode.　The　results　provide　the　government　with　a　new　idea　and　method　to　formulate　SD　policies.　©　2023　Elsevier　Ltd