I-OpenAI ithathe isinyathelo esiqhubekayo emjahweni wobuhlakani bokwenziwa ngokwembulwa kwe -GPT-5.2, imodeli yayo ethuthuke kakhulu kuze kube manje yomsebenzi wobungcweti nowebhizinisi . Lokhu kubuyekezwa kufika ngesivinini esingavamile: kuvela kungakapheli inyanga ngemuva kwe-GPT-5.1 futhi phakathi nempi yezobuchwepheshe ne-Google, i-Anthropic, kanye nabanye abancintisana nabo abasheshisa izindlela zabo zokuthuthukiswa.
Le nkampani igcizelela ukuthi le nguqulo akuyona into eyenzeka ngokuphelele esizukulwaneni, kodwa kunalokho ingukuvela kochungechunge lwe-GPT-5 olugxile ekucabangeni, ekunciphiseni amaphutha, kanye nokuxhumanisa amathuluzi. Ukwethulwa kuthinta ngqo abasebenzisi be-ChatGPT eSpain naseYurophu, lapho ukusetshenziswa komsizi sekuzinzile khona kokubili ezindaweni zemfundo kanye namabhizinisi ayihlanganisa emisebenzini yawo yansuku zonke.
Ukwethulwa ngesivinini "sekhodi ebomvu"
Isimemezelo se-GPT-5.2 siza phakathi kwendawo enokuncintisana okukhulu. Ezinsukwini ezimbalwa ngaphambili, uSam Altman wamemezela ngaphakathi " ikhodi ebomvu " ukugxila kwizinsiza ekuthuthukiseni i-ChatGPT nokubeka eceleni okwesikhashana amaphrojekthi amaningi okuhlola, njengokwethulwa kwezimemezelo noma ukuthunyelwa okukhulu kwama-ejenti azimele. Okubaluleke kakhulu, njengoba etshela abasebenzi, kwakuwukuthuthukisa ngokusobala ikhwalithi yezimpendulo kanye nokusebenza emisebenzini yobungcweti.
Ithimba lezinhlelo zokusebenza le-OpenAI licacisile ukuthi i-GPT-5.2 ayizange ikhishwe kusenesikhathi ngenxa yaleyo "khodi ebomvu ," kodwa kunalokho ibilokhu ithuthukiswa izinyanga eziningi. Noma kunjalo, inkampani ngokwayo iyavuma ukuthi amaqembu engeziwe azinikele ngqo kwi-ChatGPT abe usizo ekucwengisiseni imininingwane nasekuqedeni inguqulo ngokushesha kakhulu kune-GPT-5.1.
Ukusheshisa kuchazwa futhi yingcindezi evela ku -Gemini 3 , imodeli ye-Google eye yathola indawo emasontweni amuva nje ngabasebenzisi nasezivivinyweni zokusebenza ezahlukahlukene, ikakhulukazi ekucabangeni okuvamile. Ngesikhathi esifanayo, i-Anthropic ibilokhu iqhubekela phambili enkundleni yokuhlela ngamamodeli ayo agxile kukhodi. Kulesi simo, i-GPT-5.2 yethulwa njengempendulo ye-OpenAI yokubuyisela indawo esigabeni se-AI sezinga lobuchwepheshe.
Ezingeni lemakethe, inkampani iyavuma ukuthi ishintshe isu layo: ukugxila okuncane ekuhlolweni kwezentengiselwano njengokukhangisa kanye nokuqinisa isisekelo sobuchwepheshe se-ChatGPT kanye ne-API yayo ukuze ihlale ibalulekile kumakhasimende ebhizinisi kanye nokuphathwa komphakathi, naseYurophu.
Ukuthuthuka ngaphakathi kochungechunge lwe-GPT-5
I-OpenAI iveza i-GPT-5.2 njengokuphindaphinda okulandelayo okuqondile kwe-GPT-5.1 , hhayi njengemodeli yesizukulwane esilandelayo. Umgomo oyinhloko ukuthuthukisa umsebenzi wolwazi: ukuhlela, ukuhlaziywa kwamadokhumenti, umbono, amaspredishithi, izethulo, kanye namaphrojekthi aqukethe izinyathelo eziningi ezixhumene.
Phakathi kwezinguquko eziphawuleka kakhulu ukuphathwa okuthuthukisiwe kwezimo ezinde , okubalulekile ezinkampanini eziphatha izinkontileka ezinde, imibiko yobuchwepheshe, noma imibhalo yezomthetho eyinkimbinkimbi. Imodeli iyakwazi ukulandela intambo yezingxoxo kanye nemibhalo emikhulu, okunciphisa ukulahleka kokubumbana ezigabeni ezilandelayo.
Le nkampani iphinde ithi i-GPT-5.2 iqondisa kangcono ukusetshenziswa kwamathuluzi angaphandle (ama-API, izizindalwazi, abahumushi bekhodi, izinhlelo zangaphakathi), okuyiphuzu elibalulekile kubathuthukisi kanye neminyango ye-IT ekwakheni ama-ejenti akwazi ukwenza izenzo zomhlaba wangempela ngaphandle kokuphendula ngombhalo. Lokhu kuthuthukiswa kuqinisa ukuhlanganiswa kwezinhlelo zokusebenza zezinkampani zangaphandle emisebenzini ezenzakalelayo.
Nakuba ukwakheka okuphelele kuhlala kuhambisana ne-GPT-5, i-OpenAI ithi ilungise izinqubo zokuqeqesha kanye nezangemva kokuqeqeshwa ukuze kuncishiswe ukubona izinto ezingekho kanye namaphutha eqiniso . Ezivivinyweni zangaphakathi uma kuqhathaniswa ne-GPT-5.1, uhlobo lokucabanga oluhlukile lunciphisa cishe ingxenye yesithathu yezimpendulo ezingalungile ngokusobala emibuzweni eqondile.
Izinhlobo ezintathu: Okusheshayo, Okucabangayo, kanye Nobuchwepheshe
Njengoba kunjalo ngokukhishwa kwakamuva, i-GPT-5.2 iza ngezinguqulo ezintathu ezenzelwe ukusetshenziswa okuhlukile: Okusheshayo, Okucabangayo, kanye ne-Pro . Esikhundleni sokwengeza izici ezikhethekile, isibuyekezo siqinisa iphrofayili yenguqulo ngayinye ukuze sihlukanise kangcono ukuthi imodeli ngayinye inconyelwa ini.
I-GPT-5.2 Instant iyindlela eshesha kakhulu futhi engabizi kakhulu. Iqondiswe emibuzweni yansuku zonke, ukubhala okusheshayo, ukusesha ulwazi, kanye nokusekelwa kocwaningo . I-OpenAI ikhombisa ukuthi inikeza izimpendulo ezizinzile kakhulu kuneyangaphambili kulolu hlu, ngezincazelo ezicacile kanye nezinguquko ezimbalwa ezindinganisweni zemibuzo efanayo.
Ukucabanga kwe-GPT-5.2 kuyimodeli yokucabanga ejulile futhi yiyona ethuthuke kakhulu kulesi sizukulwane. Iqeqeshwe ukuphatha imisebenzi eyinkimbinkimbi, enezinyathelo eziningi , njengohlelo oluthuthukisiwe, ukuhlaziywa kwezezimali, ukuhumusha amadokhumenti amade, kanye nokuhlela amaphrojekthi okuningiliziwe. Iningi lentuthuko ezibalweni, kwisayensi, kanye nokuxhumanisa amathuluzi ligxile lapha.
Ekugcineni, i-GPT-5.2 Pro iphezulu kakhulu: iyimodeli enamandla kakhulu futhi ebiza kakhulu ukuyisebenzisa , eyenzelwe izindawo lapho ukunemba kubaluleke kakhulu kunesikhathi sokulibaziseka, njengocwaningo lobuchwepheshe, imisebenzi yezomthetho eyinkimbinkimbi, noma amaphrojekthi okuhlaziya idatha adingayo. Inikeza izinga eliphakeme kakhulu lemininingwane nokulawula, ngokushintshana nokusetshenziswa kwamandla ekhompyutha okukhulu.
Lokhu kuhlukaniswa kuhlose ukuvumela abasebenzisi abakhokhayo ngabanye kanye namabhizinisi kanye nabathuthukisi ukuthi bakhethe imodeli efanelana kahle nokusetshenziswa kwabo , kunokukhetha ngokungakhethi leyo ebiza kakhulu. Ezinhlanganweni zaseYurophu ezinemikhawulo yesabelomali noma yokulawula, lokhu kuhlukaniswa kungenza kube lula ukuhlela izindleko kanye nomsebenzi.
Imiphumela yokuhlolwa: kusukela kudeskithophu kuya kuma-benchmarks
I-OpenAI isekela ukuqaliswa ngenani elikhulu lezibalo zokusebenza. Ku- suite yokuhlola ye-GDPval, eklanyelwe yinkampani uqobo ukulingisa imisebenzi engu-44 yangempela, i-GPT-5.2 Thinking iyakwazi ukunqoba noma ukulingana nochwepheshe babantu cishe ku-70,9% wemisebenzi, iyiqede ngokushesha okukhulu nangezindleko zokusebenza eziphansi kakhulu.
Emkhakheni wesayensi, imodeli ithuthukisa amaphuzu ayo ku -GPQA Diamond , ukuhlolwa okulinganisa ukucabanga okusezingeni eliphezulu kwesayensi, kanye naku -FrontierMath , egxile ezinkingeni zezibalo ezithuthukisiwe. Kwezinye izinhlobo zokuhlola, njenge-AIME 2025, i-OpenAI ibika ukuthi i-GPT-5.2 ithola amaphuzu acishe abe yi-100% ezintweni ezicatshangelwayo.
Ngale kwezilinganiso, inkampani iqokomisa ukuhlolwa kwangaphakathi kwemisebenzi ejwayelekile yokuhlaziya ezezimali , njengezibonelo zezifundazwe ezintathu noma ukulingisa kokuthenga okusebenzisekayo. Kulezi zimo, kubikwa ukuthi ukucabanga kwe-GPT-5.2 kuthuthuke kusuka kumaphuzu aphakathi angaba ngu-59% kuya ngaphezu kuka-68%, kusondela ezingeni lamaqembu akhethekile.
Kuphinde kwaphawulwa ukuthuthuka ekwakhiweni kwamaspredishithi nezethulo ezinezakhiwo ezicacile kanye namaphutha ambalwa okuvumelana. Izinkampani zobuchwepheshe ezisebenza ne-OpenAI—njengezinkundla zokukhiqiza noma izinkampani ezintsha zesofthiwe zaseYurophu—kubikwa ukuthi zibone intuthuko ekucabangeni okude kanye nokusetshenziswa okuqhubekayo kwamathuluzi azenzakalelayo ngaphakathi kwemodeli.
Nokho, inkampani igcizelela ukuthi izilinganiso ziyingxenye nje yendaba nokuthi ukusebenza kwangempela kwabasebenzisi bokugcina eSpain naseYurophu kuzoncika ekutheni la makhono ahlanganiswe kanjani ezimweni ezithile zokusetshenziswa: kusukela kubasebenzi abazimele abalungiselela izabelomali noma imibiko, kuya ezinkampanini ezinkulu ezithumela abasizi bangaphakathi kubasebenzi bazo.
Amaphutha ambalwa kanye nokuzinza okwengeziwe ekusebenzeni kwansuku zonke
Omunye wemiyalezo eyinhloko ye-OpenAI ukunciphisa imibono engekho kanye namaphutha eqiniso . Ngokusho kwedatha yabo, i-GPT-5.2 Thinking ikhiqiza izimpendulo ezingalungile ezingaphansi kuka-38% kune-GPT-5.1 ezivivinyweni zangaphakathi zemibuzo eqondile. Lokhu kubaluleke kakhulu kulabo abasebenzisa imodeli ukusekela ukubika, ucwaningo, noma izinqumo zebhizinisi.
Ngaphandle kokunemba, inkampani ithi ithuthukise ikhono lomodeli lokubona amaphutha ayo ngaphambi kokunikeza impendulo yokugcina. Ngamanye amazwi, i-GPT-5.2 yenza ukuhlola kwangaphakathi okwengeziwe futhi ilahla izixazululo ezibona zingathembekile, okuholela kokuhlangenwe nakho okubikezelwayo kancane phakathi nezingxoxo ezinde.
Lokhu kubuyekezwa kuza ngemuva kwezigameko eziphikisanayo lapho i-OpenAI yagxekwa khona ngendima ye-chatbot ezingxoxweni ezibucayi , okuhlanganisa necala losemusha owazibulala ngemuva kokuyisebenzisa njengesifuba. Ngaphandle kokungena emininingwaneni yezomthetho, inkampani ithi i-GPT-5.2 iletha ukuthuthukiswa ekusingathweni kwezihloko ezibucayi, ikakhulukazi izimpawu zokucindezeleka ngokomzwelo.
Empeleni, imodeli kufanele ivumelanise kangcono ithoni yayo kanye nezincomo zayo uma ithola ukuthi isebenzelana nabasebenzisi abasengozini, yize i-OpenAI iphikelela ukuthi ayithathi indawo yokusekelwa kochwepheshe kwezokwelapha noma kwengqondo, okuyisixwayiso esifanele ngokubheka imithetho yaseYurophu nge-AI.
Indawo ekhululekile kakhulu yabathuthukisi namabhizinisi
I-GPT-5.2 ihloselwe ngqo imikhiqizo nezinsizakalo zokwakha ku-OpenAI API. Lolu hlobo lokucabanga lubonisa ukusebenza okuphezulu ekuhlolweni kokuthuthukiswa kwesofthiwe okudinga ukusebenzisa izinguquko eziphelele neziqhubekayo kumaphrojekthi wangempela, hhayi nje ukuxazulula izivivinyo ezihlukile.
Amapulatifomu okuthuthukisa ngokubambisana kanye namathuluzi okuhlela asizwayo asevele esebenza ne-GPT-5.1 abika ukuthi imodeli entsha ihlela kangcono ukulandelana kwezinyathelo eziyinkimbinkimbi , okuholela ekuphindaphindweni okuncane ngesandla ukuze kuqedwe imisebenzi efana nokulungisa ikhodi, ukusebenzisa izibuyekezo, noma ukukhiqiza amamojula kusuka ekucacisweni kombhalo.
Ezimweni zebhizinisi zaseYurophu, lapho ukuhlanganiswa nezinhlelo zangaphakathi kuvame ukuba nokugcinwa kwemithetho ngenxa yezinkinga zokuthobela imithetho, ukuthuthukiswa kokuvumelana kanye nokulawula ushintsho kungasiza ukwamukelwa kwama-ejenti enza izenzo ezizenzakalelayo ngaphandle kokuqondiswa komuntu ngqo.
Esinye isici esibalulekile ukusebenza kwemodeli ngemibhalo ebanzi yenkampani , njengoba i-GPT-5.2 ithuthukisa ukufundwa, ukufingqa, kanye nokuqhathaniswa kwezinkontileka, izinqubomgomo zangaphakathi, noma imibiko yobuchwepheshe, into ethakazelisa kakhulu izinkampani zomthetho, izinhlangano zokubonisana, kanye neminyango yokuthobela imithetho eSpain nase-EU.
Ukutholakala, ukuthunyelwa, kanye namanani
I-OpenAI isiqalile ukukhishwa kwe-GPT-5.2 emhlabeni wonke ku-ChatGPT ngezinhlelo zayo ezikhokhelwayo , okuhlanganisa i-Plus, i-Pro, i-Go, i-Business, kanye ne-Enterprise. Abasebenzisi balezi zinhlelo bazobona kancane kancane inketho yokukhetha izinhlobo ze-Instant, i-Thinking, kanye ne-Pro, kuyilapho inguqulo yamahhala izoqhubeka nokusebenzisa amamodeli angaphambilini.
Ku-API, i-GPT-5.2 Thinking ivezwa ngaphansi kwegama elithi gpt-5.2 , kuyilapho inguqulo esheshayo, egxile engxoxweni ivela njenge-gpt-5.2-chat-latest. I-OpenAI izogcina i-GPT-5.1 itholakala isikhathi sokuguquka sezinyanga eziningana, kodwa inhloso yayo eshiwoyo iwukuthuthela kancane kancane ingxenye enkulu yokusetshenziswa ochungechungeni olusha.
Ngokuphathelene namanani, i-GPT-5.2 ikhokhisa amathokheni okufaka ayizigidi ezingu-$1,75 kanye namathokheni okukhipha ayizigidi ezingu-$14 nge-API, okuphakeme kunezindleko ze-GPT-5.1. Naphezu kwalokhu, inkampani ithi ukusebenza kahle kakhulu kwemodeli emisebenzini eyinkimbinkimbi kunganciphisa izindleko eziphelele ngephrojekthi ngayinye ngokudinga izingcingo ezimbalwa kanye nokubuyekezwa.
Ezinkampanini zaseSpain nezaseYurophu ezibaluleke kakhulu ekuhlaziyweni, ekukhiqizweni kwamadokhumenti, noma ezenzakalelayo zenqubo, umqondo we-OpenAI ukuthengisa i-GPT-5.2 njengotshalomali ekuqondeni kunokuba ithuluzi elingabizi kakhulu. Lokhu kubekwa kuhambisana nesu layo lokubhekisa engxenyeni yenkampani enenani eliphezulu.
Ngokuphathelene nokwamukelwa, i-ChatGPT isalokhu ingenye yama-chatbot e-AI asetshenziswa kakhulu emhlabeni, enamakhulu ezigidi zabasebenzisi abasebenzayo masonto onke . Le nkampani iqinisekile ukuthi ukuthuthukisa ulwazi lomsebenzisi lwansuku zonke, okwenza kube kuhle kakhulu futhi kubikezelwe, kuzothuthukisa ukusetshenziswa kobungcweti, ikakhulukazi ezimakethe lapho isivele ikhona khona kakhulu, njengeSpain.
Ukuncintisana, ingqalasizinda kanye nomhlahlandlela
Ukwethulwa kwe-GPT-5.2 kuza ngesikhathi lapho imijikelezo yokuvuselelwa kwamamodeli ahamba phambili inciphile kusukela eminyakeni kuya emavikini . Lesi sivinini senza kucace ukuthi inzuzo yezobuchwepheshe iya ngokuya idlula ngokushesha nokuthi izinkampani kulo mkhakha zithola izindleko ezikhulayo ukuze zihambisane nesivinini.
Nge-Gemini 3, i-Google isebenzise ukufinyelela kwayo okuyimfihlo kudatha kanye nohlelo lwayo lokukhiqiza (Amadokhumenti, Amashidi, Idrayivu, Indawo Yokusebenzela) ukuze ihlanganise i-AI kumathuluzi asevele ekhona emahhovisi emhlabeni jikelele. Okwamanje, i-Anthropic igxile ekuphepheni nasekusebenzeni ezinhlelweni, iqinisa isikhundla sayo ezigabeni ezithile.
Ukuze iqhubeke nalesi sivinini, i-OpenAI iqinisa ingqalasizinda yayo ngezivumelwano nabakhiqizi be-chip kanye nabahlinzeki bamafu , kanye nokutshalwa kwezimali okukhulu ezikhungweni zedatha. Le nkampani iyavuma ukuthi izindleko zokusebenza kwamamodeli okucabanga athuthukile zikhuphuke kakhulu, zidlula ukwesekwa okwanikezwa ekuqaleni yizikweletu zamafu ezivela kubalingani bayo, futhi ikhomba isivumelwano ne-AWS sokusekela ingqalasizinda yayo njengenye yezinyathelo zayo ezibalulekile.
Ngesikhathi esifanayo, i-OpenAI ithembele ekubambisaneni okunesu nezinkampani ezinkulu , njenge-Disney ne -Walmart , ezibone amamodeli ayo njengendlela yokudala amafomethi amasha okuqukethwe kanye nokuhlangenwe nakho kwedijithali. Lezi zivumelwano, nakuba zingeyona ingxenye eqondile yokusebenza kwe-GPT-5.2, zibonisa ukugxila kwenkampani ekusetshenzisweni okukhiqiza imali engenayo ephindaphindayo futhi kuthethelele ukutshalwa kwezimali okukhulu kwengqalasizinda.
EYurophu, lokhu kwanda kuzodingeka kuhambisane nohlaka lokulawula oluqinile, oluphawulwe yi- EU Artificial Intelligence Regulation , oluzodinga ukucaca okukhulu ekusebenzeni kwamamodeli afana ne-GPT-5.2 kanye nasekusetshenzisweni kwezinkampani kanye nohulumeni basekhaya ngalezi zinhlelo ezinkonzweni zabo.
Nge-GPT-5.2, i-OpenAI ihlose ukuqinisa isikhundla sayo njengomhlinzeki ohamba phambili we-AI yokukhiqiza ochwepheshe nezinhlangano, iqinisa ukucabanga, ukuthembeka, kanye namakhono okuhlanganisa ngaphandle kokwethula ubuhlakani obusha . Lesi sinyathelo sifika ngesikhathi lapho imakethe iqala ukudinga imiboniso embalwa emangalisayo kanye namathuluzi azinzile axazulula izinkinga zomsebenzi wangempela, ikakhulukazi ngaphakathi kwemikhakha yebhizinisi laseSpain nelaseYurophu, lapho ukwamukelwa kwe-AI kuba yinto ebalulekile yokuncintisana.
