I-Zhipu namuhla isamemeza ukuthi i-GLM-5.3 ye-API isivulelekile ngokusemthethweni kusukela namuhla. I-model ihlakaniphile emisebenzini yezokukopisha ezinzima, ukuphepha kwezokuvikela kanye nokusebenza kwesikhathi eside.
Ubungqongqoshe bathi, i-GLM-5.3 ithole amaphuzu angu-60 ku-Artificial Analysis Intelligence Index (AA comprehensive intelligence index) emhlabeni jikelele, ingena esigabeni samamodeli aphezulu, ifana ne-Claude Fable 5, i-GPT-5.6 Sol namanye amamodeli aphezulu afihlekile, futhi ihlanganiswe ne-Kimi K3 njengomodeli wokuqala wokuvaluleka.
Ngenkathi ifinyelela ezingeni elilinganiselwe ngokwezinkululeko, i-GLM-5.3 ngokusebenzisa isayizi esincane semodeli kanye nezindleko ezingaphansi zokubiza, ihlakaniphise kakhulu ukusetshenziswa kwamamodeli aphezulu.
Ebhizini, i-Axis yenduna i-Artificial Analysis intelligence index, i-axis yesobhu i-average cost yokuphothula umsebenzi owodwa. I-GLM-5.3 ifinyelela ezindlekeni eziphansi zokubiza umsebenzi owodwa phakathi kwamamodeli aphezulu, isendaweni ephakathi kwe-high intelligence, low cost, futhi ihlanganiswe ne-“intelligence — cost” Pareto frontier yamodeli aphezulu ngokukhulu.
Ubungqongqoshe bathi, i-GLM-5.3 ayikho nje ukufisa ukuphakamisa izinkululeko, futhi izimisele ukukhulisa izinkululeko ezitholakala ngokusebenzisa izindleko ezingaphezulu. Njengoba i-API isivulelekile ngokusemthethweni, izinkululeko eziphambili zizovula ezimeni zokusebenza zabadaluli abaningi, amabhizinisi kanye nabantu abaqoka izinkululeko.
Ngokulandelana kwe-IT Weza, i-GLM-5.3 isebenzisa isisekelo esifanayo somodeli we-GLM-5.2, ukuthuthuka kuhle kakhulu kuyavela emuva kwe-training.
Manje, i-GLM-5.3 sefakiwe ema-platforms okukopisha afana ne-ZCode, futhi ifakwe i-GLM Coding Plan. I-API isivulelekile ukubizwa, inani elilinganayo elilinganayo ne-GLM-5.2. Amalungu ezinkululeko ezomodeli azovuleka ngolwesihlanu.