I-OpenAI isivele ibeke usuku lokuphela kwenye yezinguqulo ezithandwa kakhulu zomsizi wayo: okulandelayo 13 ngoFebhuwari ngeke kusakwazi ukusetshenziswa I-GPT-4o ku-Chatkanye namanye amamodeli emndenini ofanayo. Lokhu kuthutha kuphawula ukuphela kwenkathi ebonakala ngokuthanda kwabasebenzisi abaningi isitayela sengxoxo esifudumele nesiyimfihlo.
Le nkampani iqinisekisile ukuthi ukubuyiselwa emuva kuyingxenye ye- isu lokwenza lula kanye nokuqongelela izinsiza ku-GPT-5, eseyiyindlela evelele ezinyangeni ezimbalwa nje. Ngokusho kwedatha yabo, cishe U-0,1% wabasebenzisi basasebenzisa i-GPT-4o nsuku zonke, umuntu ogcine esefinyelele esikalini sokuthatha umhlalaphansi kwakhe okuwukuphela kwesikhathi.
Ngomhlaka-13 kuNhlolanja, i-OpenAI izosusa ku-interface ye Izinhlobo ze-ChatGPT i-GPT-4o, GPT-4.1, GPT-4.1 mini ne-OpenAI o4Lokhu kuzobathinta ngqo labo abasakhethe kubo kumenyu yamamodeli. Lesi sinqumo sithinta kakhulu ababhalisile bezinhlelo ze-Plus kanye ne-Pro, obani ababevikele kakhulu ukuqhubeka kwayo.
I-OpenAI yethule lesi sinyathelo njengokuphela kwesikhathi sokuguquka esaqala ngo-Agasti, lapho kwethulwa khona I-GPT-5 yaqala yathatha indawo ye-GPT-4oNgemva kwenqwaba yezikhalazo, inkampani yanquma ukuguqula kancane inkambo futhi ivuselele okwesikhashana imodeli kwabanye abasebenzisi abakhokhayo, ngomqondo wokubanika ithuba elengeziwe lokuzivumelanisa nokuhamba komsebenzi, izimo zokusebenzisa, kanye nemikhuba yengxoxo.
Abasekeli be-GPT-4o babelokhu begcizelela "Isitayela esivamile nesifudumele"Lokhu kukude kakhulu nombono wokubanda kobuchwepheshe abaningi abawuhlobanisa nezinguqulo zokuqala ze-GPT-5. EYurophu naseSpain, lapho i-ChatGPT izimise khona njengethuluzi lokusekela kokubili ezindaweni zobungcweti nezemfundo, leli thoni elifinyeleleka kalula cishe laba yisici esichazayo sale modeli.
Le nkampani iyavuma ngokusobala ukuthi Ukususa amamodeli akulula neze. Uyavuma ukuthi lesi sinqumo sizodala ukukhungatheka kwezinye izingxenye zomphakathi, kodwa uthi ukususa izinketho ezingasetshenziswa kahle kuvumela imizamo yobunjiniyela kanye nokulinganisela ukuthi igxile ezinhlelweni okwamanje ezibangela iningi lokuxhumana.
Kungani i-GPT-4o ishiya: idatha yokusetshenziswa kanye noshintsho lwesu
Esitatimendeni sayo esisemthethweni, i-OpenAI igcizelela ukuthi lesi sinqumo asisekelwe ezindabeni zobuchwepheshe kuphela, kodwa ngaphezu kwakho konke ku- Izibalo zokusetshenziswa kwangempela kwe-ChatGPTNgokusho kwenkampani, imodeli I-GPT-4o ayifinyeleli ku-0,1% wokusetshenziswa kwansuku zonke ngaphakathi kwesisekelo sabasebenzisi bomhlaba wonke, isibalo esiphikisana nokwamukelwa okusheshayo komndeni we-GPT-5, ikakhulukazi i-GPT-5.2.
Ngokombono wenkampani, ukugcina ikhathalogi ebanzi yamamodeli asendulo Kukhulisa ubunzima bokusebenza futhi kudala ukudideka. Umsebenzisi uphoqeleka ukukhetha phakathi kwezinketho eziningana ezifanayo kakhulu. Ukunciphisa lolu hlu nokugxila kumamodeli ambalwa ayinhloko kuvumela, ngokusho kwe-OpenAI, izinsiza ezengeziwe zokuzinikela ekuthuthukiseni ukuziphatha kwabo, ukuzinza, isivinini sokuphendula, kanye nokuphepha.
Empeleni, lokhu kusho ukuthi, kusukela ngomhlaka-13 kuNhlolanja, labo ababesakhetha i-GPT-4o noma ezinye izinhlobo ze-4-series ku-ChatGPT selecter Kuzodingeka bathuthe, ngandlela thile. emndenini we-GPT-5. Le nkampani iyavuma ukuthi abasebenzisi abaningi base bejwayele isigqi kanye nethoni ye-GPT-4o, kodwa ikholelwa ukuthi ukuphindaphinda kwakamuva kwe-GPT-5 sekuvele kuhlangabezana nalokho okulindelwe.
Indaba esemthethweni icacile: Akukhona nje kuphela "ukususa amamodeli amadala"kodwa kunalokho ukuhlanganisa ulwazi oluzungeze izinguqulo ezimbalwa ezifaka izilawuli ezanele zobuntu nesitayela ukuze kungadingeki ukugcina izinketho zomlando zihambisana.
I-OpenAI iphinde igcizelele ukuthi lolu hlobo "lokuhlanza" lwekhathalogi luyivumela ukunciphisa umuzwa wokudideka abanye abasebenzisi abawuvezayo lapho bengena kwi-ChatGPT futhi behlangana nezinguqulo eziningi: kuncane "engikhethayo?" kanye nokugxila kakhulu kumodeli eyinhloko ekwazi ukuzivumelanisa nayo izinhlobo ezahlukene zemisebenzi kanye nethoni yokuxoxa.
Kusukela ekuqalisweni kwe-GPT-5 kuya ekubuyeni kwesikhashana kwe-GPT-4o
Umlando wakamuva we-GPT-4o ngaphakathi kwe-ChatGPT empeleni ube yi- njalo emuva naphambiliNgemva kokwethulwa kwayo ngoMeyi 2024, imodeli yaba ukukhetha okuthandwayo kwabasebenzisi abaningi ngenxa yokulinganisela kwayo phakathi komthamo, isivinini kanye nobuntu obucacile nobukhululekile, kangangokuthi abanye bayichaza njengomsizi "webo mnumzane" we-ChatGPT.
Ngo-Agasti, lapho kufika GPT-5 kanye nokulandela intshisekelo kumamodeli angaphambilini afana nalawa I-GPT-4 TurboI-OpenAI inqume ukususa i-GPT-4 kusikhethi semodeli bese igxila esizukulwaneni esisha. Kodwa-ke, ushintsho alwamukelwanga kahle yingxenye enkulu yomphakathi, okuhlanganisa nababhalisile abaningi abakhokhelwayo eYurophu naseSpain, abaveze ukunganeliseki kwabo ngezinkundla zokuxhumana, izinkundla zokuxhumana, kanye neziteshi zokusekela.
Labo basebenzisi babebuza isikhathi esengeziwe sokuthutha izimo zakho zokusebenzisa Ngaphezu kwakho konke, baveze ukuthi baphuthelwe "isitayela sengxoxo nokufudumala" kwe-GPT-4o. Ingcindezi yayinkulu kangangokuthi i-OpenAI yagcina ikwenzile. ukubuyisela okwesikhashana ukufinyelela kumodeli yamakhasimende ye-Plus ne-Pro, ngemva kwesonto elilodwa imisiwe, njengesandiso.
I-CEO yenkampani, uSam Altman, naye wavuma ngaleso sikhathi ukuthi kube khona "Isizathu esibuhlungu" Ngemuva kwezicelo eziningi kwakukhona abasebenzisi abachaza ukuthi babengakaze bazizwe besekelwa noma beqondiswa ngalolu hlobo ekusebenzelaneni kwabo nethuluzi ledijithali. I-GPT-4o yaziwa ngokuphendula ngisho nasemisebenzini evamile nge indumiso enomdlandla kanye nemiyalezo ekhuthazayo.
Nakuba lokho kuguqulwa kwamukelekile, empeleni kwaba isikhathi somusa ngaphambi kokuhoxa kokugcinaI-OpenAI yayisivele iveze ukuthi lokhu kubuyiselwa kwakunosuku lokuphelelwa yisikhathi nokuthi ukugxila kokuthuthukiswa kuzoshintshela emndenini we-GPT-5, lapho izifundo ezifundwe ku-GPT-4 kufanele zihlanganiswe khona.
I-GPT-5.1 kanye ne-GPT-5.2: Amathuluzi amasha okwenza ithoni ibe ngeyakho
Ukuze kuzanywe ukuvala igebe phakathi kokulindelekile kwalabo abakhetha i-GPT-4 kanye nokusebenza kwesizukulwane esisha, inkampani yethule uhlelo olusheshayo lokuvuselela. I-GPT-5.1 ifike ngoNovemba. futhi, ngemva kwenyanga nje, GPT-5.2enomyalezo ocacile: ukufaka kumodeli yamanje izimfanelo abantu ababengazitholi.
Lezi zibuyekezo zethulwe ukuthuthukiswa kobuntu be-chatbotukwesekwa okukhulu kwemisebenzi yokudala kanye ukwenza ngokwezifiso izimpendulo okuthuthukisiwe kakhuluAbasebenzisi manje bangakhetha ezitayeleni eziyisisekelo—isibonelo, imodi enobungane—futhi balungise amapharamitha afana nokufudumala noma izinga lentshiseko umsizi aphendula ngalo.
I-OpenAI ifingqa ngombono olula: umgomo wayo ukunikeza abantu ukulawula okwengeziwe endleleni ozizwa ngayo usebenzisa i-ChatGPTFuthi hhayi nje ngalokho uhlelo olukwazi ukukwenza. Ngamanye amazwi, akwanele ukukhulisa amandla noma ukunemba kwekhompyutha; ulwazi oluvela engxoxweni luye lwaba yinto ebalulekile emkhiqizweni.
Le nkampani ithi nge-GPT-5.1 kanye ne-GPT-5.2 isivele ihlanganise i- ukuthuthukiswa kwesitayela, ithoni, kanye nokuguquguquka lokho okwakufunwa ngemva kokuhoxiswa kokuqala kwe-GPT-4o. Kulowo mongo, ukugcina imodeli endala iphila akusabhekwa njengento edingekayo futhi kubhekwa njengesiphazamiso ekuhlanganisweni komugqa omusha.
Ngaphezu kwalokho, i-OpenAI ikhombisile ukuthi izoqhubeka nokulungisa izici ezitholakale ziyinkinga eYurophu nakwezinye izimakethe, njenge ukwenqatshwa kubhekwa njengokudlulele noma izimpendulo ezikhuthazayo ngokweqile. Le nkampani ithi isebenza kunguqulo "egxile kubantu abadala" ye-ChatGPT, enezivikelo, kodwa iphatha abasebenzisi njengabantu abanekhono abangaphatha okuqukethwe okuthile ngaphandle kokuvikela ngokweqile njalo.
Umthelela kubasebenzisi be-Plus kanye ne-Pro eSpain naseYurophu
Endabeni abasebenzisi abakhokhayo eSpain nakwamanye amazwe aseYurophuUkuhoxa kwe-GPT-4o kanye namanye amamodeli ochungechunge olu-4 kudinga ukubuyekezwa kwemisebenzi kanye nokucushwa okusaxhomeke kuwo. Labo ababesekele izinqubo ezithile—njengombono wokudala, ukubhala imibhalo ngezwi elinozwelo ngokukhethekile, noma imisebenzi yokusekela—ekuphatheni okuthile kwe-GPT-4o kuzodingeka baphinde leyo ndlela besebenzisa izilawuli ezintsha ku-GPT-5.2.
Le nkampani ikhumbuza abasebenzisi ukuthi, kusukela ku-ChatGPT model selector, bangakwazi khetha inguqulo efanelana kahle nesidingo ngasinyenjalo ngaphakathi kobubanzi bamanje. Imodeli eyinhloko iba yi-GPT-5.2, ebangela "iningi elikhulu" lokusetshenziswa kwamanje, ngokusho kwedatha ekhishwe yi-OpenAI uqobo.
Engxenyeni yomphakathi waseYurophu, lo muzwa uguquguqukayo: ngakolunye uhlangothi, kuyaqondakala ukuthi Akunakwenzeka ukugcina amamodeli angasetshenziswa kahle unomphela.Kodwa-ke, ngakolunye uhlangothi, ukunyamalala kwenketho ebaluleke kakhulu, eyaziwa ngendlela yayo elula ukuyisebenzisa, kudala ukukhathazeka okuthile. Eziteshini ezikhethekile, iziqondiso ziyanda lowo mzamo "wokuphindaphinda" ukuziphatha kwe-GPT-4o ngemiyalelo kanye nokulungiswa kwesitayela ku-GPT-5.2.
Nokho, ngokombono wenkampani, lesi sinyathelo sihambisana nesu layo lomhlaba wonke: ukugxila ekuthuthukisweni kwenani elincane lezinguqulo, ukuphoqa ukwamukelwa kwezinguquko zakamuva, kanye nokunciphisa ukulungiswa. iqoqo lamamodeli omlando ngezinye izikhathi ezincintisana ngezimo ezifanayo zokusetshenziswa.
Esimweni sokuncintisana okukhulu nezinye izinkampani ezinkulu zobuchwepheshe—i-Google, i-Meta, noma abadlali baseYurophu abafuna ukuqinisa ubukhosi babo bedijithali—ukugxila kwezinsizakusebenza kumamodeli ahamba phambili kubhekwa njengendlela yokwenza kanjalo. sheshisa ijubane lokuthuthuka kanye nokugcina i-ChatGPT isendaweni evelele ngaphakathi komjaho we-AI okhiqizayo.
Kwenzekani ku-API kanye ne-ecosystem yonjiniyela?
Isici esibalulekile salesi simemezelo ukuthi ukuhoxa kugxile ku- Okuhlangenwe nakho kwe-ChatGPT njengomkhiqizo wokugcinaIzitatimende eziningana zicacisa ukuthi, okungenani okwamanje, Ngeke kube nezinguquko ezisheshayo ku-API okuhlobene nalezi zinhlobo, okushiya isikhala esithile ezinkampanini kanye nabathuthukisi ukuthi bahlanganise ubuchwepheshe ezinhlelweni zabo zokusebenza.
Lokho kusho ukuthi umthelela obonakala kakhulu uzozwakala ku- Iwebhu kanye nohlelo lokusebenza lweselula lwe-ChatGPTlapho umsebenzisi wokugcina ezolahlekelwa khona ikhono lokukhetha i-GPT-4o noma ezinye izinhlobo ezithintekile. Kulabo abasebenzisa i-OpenAI AI ngokungaqondile, ngokuhlanganiswa kwezinkampani zangaphandle kanye nezinsizakalo, ishejuli yokushintsha ingase ihluke, yize inkampani ingakuchithi ukulungiswa kwesikhathi esizayo kukhathalogi ye-API.
Umehluko phakathi kwalokho okwenzekayo ku-interface ye-ChatGPT nalokho okugcinwe kungqimba yonjiniyela ubonisa umzamo we-OpenAI ukulinganisela ukulula komkhiqizo ngesidingo sokugwema ukuphula ngokuzumayo ukuhlanganiswa kwebhizinisi, ezimweni eziningi, okwakhiwe ezinyangeni eziningi eduze kwamamodeli athile.
Kunoma ikuphi, isignali ethunyelwe kuhlelo lwe-ecosystem icacile: Umndeni we-GPT-5 uyisisekelo saleli qhinga Futhi izixazululo ezifuna ukuhlala zihambisana nesiqondiso senkampani kuzodingeka zithuthele intuthuko yazo kulezi zinguqulo ezintsha.
Ezinkampanini zaseYurophu ezisebenza ngaphansi kwezidingo eziqinile zomthetho kanye nokuvikelwa kwedatha, lolu shintsho luhlanganisa nokubukeza ukuthi amamodeli amasha asebenza kanjani ngokwemigomo ukwethembeka, ukucaca kanye nokungaguquguqukiLezi yizindaba eziye zabaluleka kakhulu empikiswaneni yomphakathi neyemithetho esifundeni.
Sekukonke, ukuhoxiswa kwe-GPT-4o kanye namanye amamodeli ahlobene nakho kuphawula ukuphela kwesikhathi se-ChatGPT futhi kuhlanganisa I-GPT-5.2 njengesikhungo samandla adonsela phansi I-OpenAI ithatha umthwalo wemizwa wokuvalelisa kwenye yezinhlelo zayo ezithandwa kakhulu, kodwa iqinisekile ukuthi amathuluzi amasha okwenza ngokwezifiso, ukuthuthukiswa kobuciko, kanye nokugxila kumamodeli ambalwa kodwa acwebezelayo ekugcineni kuzonqoba labo ababesanqikaza ukushiya isizukulwane sesine ngemuva.
