{"id":48882,"date":"2026-04-02T14:42:21","date_gmt":"2026-04-02T14:42:21","guid":{"rendered":"https:\/\/mavikumsal.com\/?p=48882"},"modified":"2026-04-17T13:37:38","modified_gmt":"2026-04-17T13:37:38","slug":"openai-gpt-5-4u-piyasaya-surdu","status":"publish","type":"post","link":"https:\/\/mavikumsal.com\/index.php\/2026\/04\/02\/openai-gpt-5-4u-piyasaya-surdu\/","title":{"rendered":"OpenAI, GPT-5.4’\u00fc piyasaya s\u00fcrd\u00fc."},"content":{"rendered":"\n

OpenAI<\/a>\u00a0, profesyonel g\u00f6revleri \u00e7\u00f6zmek i\u00e7in en g\u00fc\u00e7l\u00fc modeli olan GPT-5.4’\u00fc tan\u0131tt\u0131. Yeni geli\u015ftirme, daha \u00f6nce Codex serisi modellerinde uygulanan mant\u0131ksal d\u00fc\u015f\u00fcnme, programlama ve ajan tabanl\u0131 i\u015f ak\u0131\u015flar\u0131ndaki en son geli\u015fmeleri bir araya getiriyor ve bunlar\u0131 genel ama\u00e7l\u0131 bir modelde somutla\u015ft\u0131r\u0131yor. \u00d6zellikle elektronik tablolar, sunumlar ve belgeler olu\u015fturmak gibi i\u015f g\u00f6revleriyle \u00e7al\u0131\u015f\u0131rken verimlili\u011fe \u00f6nem veriliyor. OpenAI’ye g\u00f6re, GPT-5.4, \u00f6nceki s\u00fcr\u00fcmlere g\u00f6re daha do\u011fru, verimli ve daha az kaynak t\u00fcketimiyle karma\u015f\u0131k i\u015flemleri ger\u00e7ekle\u015ftirebiliyor.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

Model \u00fc\u00e7 ortamda mevcuttur: ChatGPT,\u00a0
Codex<\/a>\u00a0ve API arac\u0131l\u0131\u011f\u0131yla iki s\u00fcr\u00fcm. ChatGPT’de GPT-5.4 Thinking ve GPT-5.4 Pro olarak sunulmaktad\u0131r. Bir sorgu i\u015flenirken, kullan\u0131c\u0131 modelin ak\u0131l y\u00fcr\u00fctme plan\u0131n\u0131n \u00f6n tasla\u011f\u0131n\u0131 al\u0131r; bu da yan\u0131t olu\u015fturma s\u00fcrecinde do\u011frudan ayarlamalar yap\u0131lmas\u0131na olanak tan\u0131r. Ayr\u0131ca OpenAI, GPT-5.4’\u00fcn derin web arama sorgular\u0131n\u0131 ger\u00e7ekle\u015ftirme konusunda GPT-5.3 Instant’tan daha iyi performans g\u00f6sterdi\u011fini belirtmi\u015ftir.<\/p>\n\n\n\n

Teknik iyile\u015ftirmeler aras\u0131nda, bilgisayar kullan\u0131m yeteneklerindeki \u00f6nemli geli\u015fmeleri vurgulamak gerekir: model, OSWorld-Verified Computer Use k\u0131yaslamas\u0131nda %75’lik bir puan elde etti. OpenAI serisinde ilk kez, \u00f6nizleme modunda 1 milyon token hacmine sahip bir ba\u011flam penceresi deste\u011fi sunuldu. GDPval’da model, %83’l\u00fck bir puanla yeni bir rekor k\u0131rd\u0131. MMMU-Pro k\u0131yaslamas\u0131nda, GPT-5.4 harici ara\u00e7lar kullan\u0131lmadan %81,2’lik bir ba\u015far\u0131 oran\u0131 g\u00f6sterdi. Bu,\u00a0
%80,5’lik bir ba\u015far\u0131 oran\u0131 g\u00f6steren\u00a0
Gemini 3.1 Pro’dan daha y\u00fcksektir.<\/p>\n\n\n\n

\"\"<\/figure>\n\n\n\n

GPT<\/a>\u00a0modelleriyle ilgili s\u00fcrekli \u015fikayetlerden biri,\u00a0\u00fcrettikleri belgelerin ve sunumlar\u0131n yetersiz kalitesiydi. OpenAI bu eksikli\u011fi gidermek i\u00e7in \u00e7ok \u00e7al\u0131\u015ft\u0131: \u015firkete g\u00f6re, GPT-5.4 daha estetik materyaller \u00fcretiyor, daha fazla g\u00f6rsel \u00e7e\u015fitlilik sunuyor ve g\u00f6r\u00fcnt\u00fc olu\u015fturma yeteneklerinden daha iyi yararlan\u0131yor. Ayn\u0131 zamanda \u015firket, yeni modelin, belirte\u00e7 kullan\u0131m\u0131 a\u00e7\u0131s\u0131ndan t\u00fcm mant\u0131ksal \u00e7\u0131kar\u0131m modelleri aras\u0131nda en verimlisi oldu\u011funu vurguluyor.<\/p>\n\n\n\n

Kar\u015f\u0131la\u015ft\u0131rma tablolar\u0131na g\u00f6re, GPT-5.4 serisi,\u00a0Anthropic<\/a>\u00a0ve Google’\u0131n rakip modellerine k\u0131yasla \u00e7o\u011fu temel g\u00f6stergede daha iyi performans g\u00f6stermektedir.<\/p>\n\n\n\n

Geli\u015ftiriciler i\u00e7in Codex ortam\u0131nda, GPT-5.4 kullan\u0131l\u0131rken sonu\u00e7 kalitesinden \u00f6d\u00fcn vermeden token i\u015fleme h\u0131z\u0131n\u0131 1,5 kata kadar art\u0131ran yeni bir \/fast modu bulunmaktad\u0131r. API’de, istekte t\u00fcm ara\u00e7 tan\u0131mlar\u0131n\u0131 \u00f6nceden y\u00fcklemek yerine, yeni model, arama fonksiyonuyla birlikte mevcut ara\u00e7lar\u0131n hafif bir listesini al\u0131r.<\/p>\n\n\n\n

GPT-5.4 Thinking, ChatGPT Plus, Team ve Pro kullan\u0131c\u0131lar\u0131 i\u00e7in zaten mevcut olup GPT-5.2 Thinking’in yerini alm\u0131\u015ft\u0131r. Ayn\u0131 zamanda, GPT-5.2 Thinking \u00fc\u00e7 ay daha eski modeller listesinde kalacakt\u0131r. GPT-5.4 Pro, yaln\u0131zca ChatGPT Pro ve Enterprise planlar\u0131n\u0131n aboneleri i\u00e7in tasarlanm\u0131\u015ft\u0131r. API fiyatlar\u0131na gelince: GPT-5.4, gelen token ba\u015f\u0131na 1 milyon i\u00e7in 2,50 ABD dolar\u0131 ve giden token ba\u015f\u0131na 1 milyon i\u00e7in 15 ABD dolar\u0131 tutar\u0131ndad\u0131r. GPT-5.4-pro s\u00fcr\u00fcm\u00fc \u00e7ok daha pahal\u0131d\u0131r \u2013 gelen token ba\u015f\u0131na 1 milyon i\u00e7in 30 ABD dolar\u0131 ve giden token ba\u015f\u0131na 1 milyon i\u00e7in 180 ABD dolar\u0131.<\/p>\n","protected":false},"excerpt":{"rendered":"

OpenAI, profesyonel g\u00f6revler i\u00e7in yeni bir \u00e7\u0131\u011f\u0131r a\u00e7an model olan GPT-5.4’\u00fc piyasaya s\u00fcrd\u00fc.<\/p>\n","protected":false},"author":1,"featured_media":48883,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"tdm_status":"","tdm_grid_status":"","footnotes":""},"categories":[1],"tags":[119,82],"class_list":["post-48882","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-genel","tag-gpt-5-4","tag-openai"],"_links":{"self":[{"href":"https:\/\/mavikumsal.com\/index.php\/wp-json\/wp\/v2\/posts\/48882","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/mavikumsal.com\/index.php\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/mavikumsal.com\/index.php\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/mavikumsal.com\/index.php\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/mavikumsal.com\/index.php\/wp-json\/wp\/v2\/comments?post=48882"}],"version-history":[{"count":1,"href":"https:\/\/mavikumsal.com\/index.php\/wp-json\/wp\/v2\/posts\/48882\/revisions"}],"predecessor-version":[{"id":48888,"href":"https:\/\/mavikumsal.com\/index.php\/wp-json\/wp\/v2\/posts\/48882\/revisions\/48888"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/mavikumsal.com\/index.php\/wp-json\/wp\/v2\/media\/48883"}],"wp:attachment":[{"href":"https:\/\/mavikumsal.com\/index.php\/wp-json\/wp\/v2\/media?parent=48882"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/mavikumsal.com\/index.php\/wp-json\/wp\/v2\/categories?post=48882"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/mavikumsal.com\/index.php\/wp-json\/wp\/v2\/tags?post=48882"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}