Ukwethulwa kwe -Gemini 3 kaGoogle kushintshe ijubane lomncintiswano wobuhlakani bokwenziwa obukhiqizayo. Ngemva kweminyaka eminingi idlula igagasi lomthelela wokuqala we-ChatGPT, inkampani isheshise kokubili ukuthuthukiswa kanye nokusetshenziswa komodeli wayo omusha, kanye namaphuzu okuqala edatha afinyelela eqophelweni lokuguquka embonini.
Emasontweni ambalwa nje, i-Gemini 3, i-Gemini 3 Pro, kanye ne-Gemini 3 Deep Think zisuke ekutholakaleni okulinganiselwe e-United States zaya ezimakethe eziningi, zinobukhona obuqinile eYurophu naseSpain. Ngesikhathi esifanayo, ukusebenza kwazo ezivivinyweni zokulinganisa kanye nokuthandwa kwazo phakathi kwabasebenzisi abathuthukile kukhulise ingcindezi ku-OpenAI, ephendule ngokusebenzisa "isaziso esibomvu" sangaphakathi ukuqinisa i-ChatGPT.
Ukukhishwa okusheshisiwe emhlabeni jikelele kwe-Gemini 3 futhi kugxile e-Europe
Kulokhu, i-Google ikhethe indlela esheshayo kakhulu kunasekuqalisweni kwangaphambilini: i-Gemini 3 ifakwa kwimodi ye-AI yenjini yokusesha emazweni angaphezu kuka-120 kulo lonke elaseMelika, e-Asia, kanye nesifunda se-EMEA. Lokhu kukhishwa okukhulu kuhluke kakhulu nesu elihamba kancane kancane inkampani elilandele kumamodeli angaphambilini.
Ngaphakathi kwale mephu, ukwanda eYurophu kuphawuleka kakhulu . Izimakethe ezifana neSpain, iFrance, iJalimane, i-Italy, i-United Kingdom, kanye namazwe aseNordic sezingakwazi kakade ukufinyelela imodeli entsha ngemodi ye-AI ye-Google Search. Amazwe amaningana eNyakatho Afrika, eMpumalanga Ephakathi, kanye nase-Afrika engaphansi kweSahara nawo ayajoyina, okudala ukuba khona okubanzi kakhulu esifundeni.
Kulezi zifunda, abasebenzisi ababhalisele i-Google AI Pro kanye ne-Google AI Ultra bangakhetha inketho ethi "Ukucabanga nge-3 Pro" ngaphakathi kwemodi ye-AI yenjini yokusesha ukuze baqalise imibuzo yabo besebenzisa imodeli ethuthukisiwe kakhulu, enenjongo evamile. Kodwa-ke, okwamanje, i-Gemini 3 Pro itholakala kuphela ngesiNgisi , imininingwane efanele emazweni afana neSpain, lapho ingxenye enkulu yabantu ingakajwayelani namamodeli e-AI ngalolo limi.
Le nkampani igcizelela ukuthi i-Gemini 3 ithuthukisa ukuqonda inhloso yomsebenzisi kanye nokufunda okusheshayo ngenxa yenjini yokucabanga evuselelwe. Umgomo ukunciphisa isidingo semiyalelo enemininingwane eminingi nokunikeza izimpendulo ezinembile kakhulu ekuphindaphindweni okuncane, kokubili emisebenzini yansuku zonke yemibuzo kanye nezimo zobungcweti.
Ngesikhathi esifanayo, i-Google yandisa i-Nano Banana Pro , ithuluzi layo lokukhiqiza izithombe nokuhlela elihlanganiswe emndenini we-Gemini 3, emazweni amaningi. Le ngxenye ebonakalayo ihlangana namakhono ombhalo nekhodi wemodeli eyinhloko ukudala uhlelo oluphelele kakhulu oluzungeze ukukhishwa okusha.
I-Gemini 3 Cabanga Ngokujulile: ukucabanga okujulile kanye nokuphendula okunensayo kodwa okunembile
Ngaphandle kwemodeli yenhloso ejwayelekile, esinye sezici ezintsha ezinkulu yi- Gemini 3 Deep Think , eyethulwe yi-Google "njengemodeli yayo yokucabanga ethuthuke kakhulu" kuze kube manje. Lolu hlobo seluvele lwethulwe eSpain nasemhlabeni jikelele kubabhalisile be-Google AI Ultra, olutholakala ngohlelo lokusebenza lwe-Gemini ngesikhethi semodeli kanye namathuluzi "oKucabanga".
I-Deep Think ivelele ngokuthembela kwayo endleleni yokucabanga ethuthukisiwe ehambisanayo , eyenzelwe ukuhlola imibono eminingi ngesikhathi esisodwa ngaphambi kokubuyisela impendulo. Le ndlela ibeka phambili ikhwalithi kunokushesha, okusobala eqinisweni lokuthi izimpendulo zingathatha imizuzu embalwa ukwenziwa lapho umsebenzi uyinkimbinkimbi.
Indlela yenkampani ihilela ukuguqula i-Gemini 3 Deep Think ibe uhlobo "lobuchopho belebhu" olugxile ezinkingeni zezibalo, isayensi, kanye ne-logic , kanye nokuhlela okuthuthukisiwe. Ngokusebenzisa imijikelezo yokucabanga ephindaphindayo, imodeli ithembisa ukuboniswa okungcono, ama-prototypes acwebezelayo, kanye nekhodi enezici eziningi kunezinguqulo zangaphambilini.
Ngokombono wobuchwepheshe, i-Deep Think yakhelwe ku- Gemini 2.5 futhi izuza ingxenye yokwakheka kwayo, kodwa yengeza uhlelo lokucabanga oluyinkimbinkimbi kakhulu. I-Google igcizelela ukuthi le ndlela ivumele intuthuko enkulu ezilinganisweni ezahlukene, ikakhulukazi lezo ezilinganisa ukucabanga okungaqondakali kanye nokusebenza kwekhodi.
Ngokuphathelene nokufinyelela, i-Gemini 3 Deep Think itholakala kuphela ngokubhaliselwe kwe-Google AI Ultra , okubiza kakhulu ohlwini. Inkampani ngokwayo iyavuma ukuthi iyimodeli edinga kakhulu ukubalwa kwezibalo edinga isikhathi eside ukuhlola imibono ehlukene ngesikhathi esifanayo. Lesi sidingo sezinsiza sichaza ukuthi kungani sinikezwa kuphela ngohlelo olubiza kakhulu kanye nokulawula ukusetshenziswa okuqinile.
Imiphumela yebhentshimakhi: I-Gemini 3 kanye ne-Deep Think iyahola
Esinye sezimpikiswano eziyinhloko ze-Google zokusekela ukufika kwe-Gemini 3, ikakhulukazi i-Gemini 3 Deep Think , imiphumela yochungechunge lwezivivinyo zokulinganisa ezizimele. Le nkampani icaphuna ngokugqamile i- Humanities' Last Exam , isivivinyo semibuzo engu-2.500 esihlanganisa izibalo, isayensi, umlando, kanye nokucabanga okuvamile.
Kulolu vivinyo, imodeli ejwayelekile ye-Gemini 3 ithole amaphuzu angu-37,5% ngaphandle kokufinyelela amathuluzi angaphandle , kuyilapho i-Deep Think entsha iphakamise leso sibalo saya ku- 41% , futhi ngaphandle kokuthembela ezinsizeni ezengeziwe. I-Google ihumusha le datha njengophawu lokuthi imodeli iyakwazi ukuxazulula izinkinga eziyinkimbinkimbi ngezinga eliphakeme lokuthembeka kunezinhlelo zangaphambilini.
Ukuqhathaniswa nezimbangi zayo eziyinhloko kungenye yamaphuzu abucayi kakhulu. Ku-Humanity's Last Exam efanayo, i-OpenAI's GPT-5.1 ithole amaphuzu angu-26,5% kuphela ngaphandle kwamathuluzi , okusobala ukuthi ingaphansi kwamamodeli amabili emndenini we-Gemini 3. Kwezinye izilinganiso, njenge- GPQA Diamond (egxile olwazini lwesayensi oluthuthukisiwe), i-Deep Think isondela ku-93,8%, kanti ku -ARC-AGI-2 (ukucabanga okubonakalayo kanye nokuxazulula amaphazili ngokusebenzisa ikhodi) ifinyelela ku- 45,1% , okuchazwe yi-Google ngokuthi "okungakaze kubonwe ngaphambili."
Kulolu vivinyo lwakamuva, umehluko uma uqhathaniswa namanye amamodeli uyamangalisa: UClaude Sonnet uthole amaphuzu angu-4.5 acishe abe ngu-13,6%, i-GPT-5.1 ifinyelela ku-17,6%, kanti i- Gemini 3 Pro ngokwayo ayitholakali ku-31,1%. Lesi sikhala sibonisa ukuthi indlela yokucabanga ye-Deep Think efanayo idala inzuzo ethile emisebenzini edinga ukuhlola ukuhlukahluka okuningi ngaphambi kokufika esixazululweni esifanele.
I-Google iphinde igqamise impumelelo yowandulelayo, i-Gemini 2.5 Deep Think , esanda kuwina indondo yegolide ku-International Mathematical Olympiad futhi yafinyelela kowamanqamu we-International Collegiate Programming Contest. Ngale ndlela, inkampani izama ukulandelela umugqa ocacile wentuthuko ekugcineni eholela esizukulwaneni sesithathu somndeni wakwaGemini.
Ngaphandle kwezibalo zelebhu, abanye abacwaningi kanye nochwepheshe be-AI bachaze i-Gemini 3 njengentuthuko ebaluleke kakhulu yonyaka kumamodeli ezilimi ezinkulu. Ikhono layo lokusebenza ngaphandle kokuphazamiseka ngombhalo, ividiyo, umsindo, kanye nekhodi liphawuleka kakhulu—indlela abaningi abayibona njengesinyathelo esibheke ezinhlelweni ezikwazi ukusingatha imisebenzi eminingi ngaphansi kwesambulela esisodwa.
Impumelelo ekusebenziseni, imikhawulo enguqulweni yamahhala, nokwanda kokusesha
Intshisekelo enamandla ku-Gemini 3 ayibonakali nje kuphela ezilinganisweni zokulinganisa kodwa nasekusetshenzisweni kwangempela. Ngokusho kwedatha eyabiwe yi-Google, isisekelo sabasebenzisi abasebenzayo benyanga zonke sohlelo lokusebenza lwe-Gemini sakhula sisuka cishe ezigidini ezingu-450 ngoJulayi saya ezigidini ezingu-650 ngo-Okthoba . Ingxenye yalokhu kukhula ihlobene ngqo nokwethulwa kwe-Gemini 3 kanye nokuhlanganiswa kwayo okujulile nezinsizakalo kulo lonke uhlelo lwenkampani.
Lokhu kwanda kubonakala nasezindleleni zokusesha. Embikweni waminyaka yonke othi " Unyaka Wokusesha ", igama elithi Gemini livela njengombuzo odumile kakhulu emhlabeni jikelele, lidlula ezinye izihloko zemidiya nemicimbi yezemidlalo. I-Google ithi lokhu kwanda, ikakhulukazi kusukela ngoSepthemba, kubangelwa ukumenyezelwa kwezici eziningi ezintsha ze-AI emikhiqizweni efana ne -Chrome, i-Search, ne-Android , lapho i-Gemini 3 ithatha indawo ephambili.
Uhlangothi olungakhangi kangako lwale mpumelelo ukuthi i-Google isiqalile ukukhawulela ukufinyelela kwezinye izici zekhwalithi ephezulu zama-akhawunti wamahhala. Ekhasini layo lokusekela, inkampani iyavuma ukuthi, ukuze ilondoloze ukuzinza kwensizakalo, kuye kwadingeka inciphise inani lemiyalelo nezingxoxo ezingathunyelwa, kanye nekhwalithi kanye nokuvama kokukhiqizwa kwezithombe kusetshenziswa i -Nano Banana Pro kwabangabhalisile.
Umqondo ongemuva kwalesi silinganiso ulula: ukuqaliswa kwe- Gemini 3 Pro kanye nokuhlangenwe nakho okuhle kokuqala kuholele ekwandeni okukhulu kokusetshenziswa, kwaze kwaba yilapho, ngaphandle kwemingcele, ukugcwala kanye nokuhlangenwe nakho komsebenzisi okonakele kungavela. Umlayezo we-Google ushiya umnyango uvulekele ukukhawulela, ngezikhathi ezithile, inani lezicelo kanye nokusetshenziswa kwezinye izici ezithuthukisiwe.
Kulabo abathole ukuthi i-Gemini 3 Pro iyahlangabezana nezidingo zabo futhi bafuna ukugwema le mikhawulo, inkampani inikeza izinhlelo zayo ezikhokhelwayo. I-Google AI Pro , enentengo engu-$19,99 ngenyanga, inikeza ukufinyelela kumodeli yakamuva yokusetshenziswa kakhulu, kanye nokuhlanganisa amathuluzi okubhala ikhodi, umsizi we-Gemini ku-Chrome, kanye ne-AI ngaphakathi kwezinhlelo zokusebenza ezifana ne-Gmail, Amadokhumenti, kanye ne-Drive . Ngaphezu kwalokho, uhlelo lwe -Google AI Ultra , olucishe lube ngu-$249,99 ngenyanga, lwandisa kakhulu amakhredithi okudala amavidiyo e-AI futhi lufaka izinzuzo ezengeziwe, luzibeka njengomnikelo owenzelwe ochwepheshe namabhizinisi anezidingo ezidinga kakhulu.
Ukuhlanganiswa ku-Google ecosystem nezindlela ezintsha zokuyisebenzisa
Esinye sezici ezibaluleke kakhulu eYurophu naseSpain ukuthi i-Gemini 3 ihlangana kanjani nokuphila kwedijithali kwansuku zonke kwabasebenzisi . Ngokungafani namanye amamodeli athembele kuzinhlelo zokusebenza ezithile noma ukuhlanganiswa kwezinkampani zangaphandle, indlela ye-Google ihilela ukuletha i-AI ngqo lapho abantu bekhona kakade.
Okokuqala, i-Gemini 3 ihlanganiswe ne-Google Search nge- "AI Mode" yomdabu. Ngaphandle kwesidingo sokulanda noma yini noma ukushintsha amapulatifomu, ababhalisile bangamane bathinte le modi futhi bafinyelele imodeli ngamakhono okucabanga athuthukile. Empeleni, lokhu kuguqula usesho lwendabuko lube yinto exoxa kakhulu futhi egxile emisebenzini.
Okwesibili, imodeli ixhumana nezinhlelo zokusebenza ze-Google Workspace . Amathuluzi afana ne-Gmail, Docs, Sheets, Drive, kanye neKhalenda angasebenzisa i-Gemini 3 ukubhala ama-imeyili, ukufingqa amadokhumenti amade, ukukhiqiza amaspredishithi ayinkimbinkimbi, noma ukusiza ekuhleleni amaphrojekthi. Ezinkampanini eziningi zaseYurophu, ikakhulukazi ama-SME asevele esungulwe ohlelweni lwe-Google, lokhu kuhlanganiswa kungaba yisizathu esibalulekile sokuzama noma ukwamukela le nsizakalo.
Ngasohlangothini lokuthuthukiswa, i-Gemini 3 iyatholakala nge-Google AI Studio kanye ne-Google Cloud , okwenza kube lula ngamathimba obuchwepheshe eSpain nakwamanye amazwe aseYurophu ukufaka imodeli ezixazululweni zawo. Kusukela kuma-chatbot ezinkampani kuya kubasizi bokuhlaziya idatha yangaphakathi, imodeli ingaba yingxenye ebalulekile yemikhiqizo emisha yedijithali.
I-Google igcizelela ukuthi i-Gemini 3 idinga ukufaka okuncane kunamamodeli angaphambilini ukuze iphile imibono yabasebenzisi, ithembele ekucabangeni kwayo okuseqophelweni eliphezulu kanye namakhono ayo amaningi ngokugcwele . Lokhu kuhlanganiswa kokuhlanganiswa okujulile, ukucabanga okuthuthukisiwe, kanye nokusekelwa kombhalo, izithombe, umsindo, ividiyo, kanye nekhodi kubalulekile esu lenkampani.
Umthelela wokuncintisana: ingcindezi ku-OpenAI kanye nomjaho wobuholi
Ukusebenza kwe -Gemini 3 akugcini nje ngokuba nemiphumela yobuchwepheshe, kodwa futhi kushintsha isimo sokuncintisana se-AI ekhiqizayo. Eminyakeni emibili yokuqala ngemva kokufika kwe-ChatGPT, i-OpenAI igcine inzuzo ecacile ngekhwalithi yemodeli ebonwayo kanye nokuba khona komphakathi. Manje, insimu iqinile kakhulu.
Ukusabela kwe-OpenAI kube okumangalisayo kakhulu. Ngokusho kwemibiko eshicilelwe yimithombo yezindaba eminingana, uSam Altman, i-CEO yenkampani, umemezele "ikhodi ebomvu" yangaphakathi ngemuva kokwethulwa kweGemini 3. Ememethweni eya kubasebenzi, kubikwa ukuthi ufune ukuthi imizamo igxile ekuthuthukiseni i-ChatGPT nokuthi ezinye izinhlelo zamasu zimiswe, echaza intuthuko ye-Google njengosongo olukhulu.
Lo "khodi ebomvu" usho ukuqondisa kabusha izinsiza kanye nokuhlehlisa amaphrojekthi ayebhekwa njengabalulekile ngaphambilini, okuhlanganisa ama-ejenti akhethekile ezindaweni ezifana nokunakekelwa kwempilo kanye nezentengiselwano, umsizi womuntu siqu othuthukile, kanye nezinhlelo zokwenza imali zokukhangisa. Umyalezo ucacile: okubaluleke kakhulu ukuqinisa i-ChatGPT nokugcina ithuba layo lokuncintisana ngokumelene namamodeli afana ne-Gemini 3 kanye ne-Deep Think.
Abahlaziyi bezimboni bakhomba ukuthi i-Google iqale kahle kakhulu ekusebenzeni kwamandla okusebenzisa kanye neziteshi zokusabalalisa , ngenxa yenethiwekhi yayo yezikhungo zedatha kanye nokuhlanganiswa okuqondile kwe-Gemini emikhiqizweni evamile njengenjini yayo yokusesha, i-Android, kanye ne-Workspace, yize abanye abadlali abanjengo- Alibaba beqinisa iminikelo yabo yamafu nge-AI . I-OpenAI, yona, ithembele kakhulu ekwenzeni imali ngamamodeli ayo kanye nokusekelwa kwabalingani abanjengoMicrosoft ukuze iqhubeke nokutshalwa kwezimali kwayo kwengqalasizinda.
Noma kunjalo, impi ayisekho. Nakuba ezinye izibikezelo zisikisela ukuthi i-Google ingagcina imodeli yayo evelele esikhathini esifushane , kugcizelelwe nokuthi i-OpenAI inesisekelo esikhulu sabasebenzisi kanye nekhono elinamandla lokusungula izinto ezintsha. U-Altman ngokwakhe ukhombisile ukuthi imodeli entsha yokucabanga iseduze, eyenzelwe ngqo ukubuyisa indawo ngokumelene ne-Gemini 3.
Ngokusatshalaliswa okusheshayo kwe -Gemini 3, ukufika kwe-Deep Think, kanye nokuhlanganiswa kwala mathuluzi empilweni yedijithali yansuku zonke , i-Google ikwazile ukubuyisa indawo elahlekile futhi yashintsha ingxoxo mayelana ne-AI yayo. Ukuhlanganiswa kwemiphumela eqinile yokulinganisa, ukuthambekela kokusesha okuphezulu, kanye nokugxila kakhulu eYurophu naseSpain kukhomba esigabeni esisha sokuncintisana okukhulu, lapho i-Google ne-OpenAI zombili ziphikisana hhayi nje kuphela ngesihloko "semodeli engcono kakhulu" kodwa futhi nesikhundla esiphezulu esizukulwaneni esilandelayo sezinsizakalo zedijithali.
