{"id":2782,"date":"2024-07-19T10:40:31","date_gmt":"2024-07-19T10:40:31","guid":{"rendered":"https:\/\/howtogeek.blog\/lt\/?p=2782"},"modified":"2024-07-19T10:40:31","modified_gmt":"2024-07-19T10:40:31","slug":"openai-unveils-gpt-4o-mini-its-cheapest-small-ai-model-yet-lt","status":"publish","type":"post","link":"https:\/\/howtogeek.blog\/lt\/openai-unveils-gpt-4o-mini-its-cheapest-small-ai-model-yet-lt\/","title":{"rendered":"OpenAI pristato GPT-4o Mini, kol kas pigiausi\u0105 ma\u017e\u0105 AI model\u012f"},"content":{"rendered":"

Kitas AI modelis prisijung\u0117 prie AI modeli\u0173 j\u016bros, o \u0161is v\u0117l yra i\u0161 OpenAI. OpenAI anks\u010diau \u0161iais metais i\u0161leido GPT-4o, kuris tur\u0117jo b\u016bti efektyvesnis nei GPT-4. Ta\u010diau tai vis tiek buvo gana brangu ir gal\u0117jo pareikalauti nema\u017eai s\u0105skait\u0173, ypa\u010d k\u016br\u0117jams, kuriems reikia pakartotinai per dien\u0105 i\u0161kviesti AI model\u012f per API savo programoms.<\/p>\n

Tod\u0117l k\u016br\u0117jai \u0117m\u0117 ie\u0161koti pigesni\u0173 konkurent\u0173 AI modeli\u0173, toki\u0173 kaip Gemini 1.5 Flash ar Claude 3 Haiku.<\/p>\n

Dabar \u201eOpenAI\u201c i\u0161leid\u017eia GPT-4o mini \u2013 ekonomi\u0161kiausi\u0105 j\u0173 model\u012f, su kuriuo jie taip pat patenka \u012f ma\u017e\u0105 AI modeli\u0173 erdv\u0119. Nors GPT-4o mini kol kas yra pigiausias j\u0173 modelis, jis nepasiekia tokios ma\u017eos kainos suma\u017eindamas intelekt\u0105; jis yra protingesnis nei j\u0173 turimas GPT-3.5 Turbo modelis.<\/p>\n

Remiantis OpenAI, GPT-4o mini surinko 82% MMLU (Measuring Massive Multitask Language Understanding), pranokdamas daugel\u012f modeli\u0173; tai yra atitinkami skirting\u0173 modeli\u0173 balai: GPT-3.5 Turbo (70 %), Claude 3 Haiku (75,2 %) ir Gemini 1.5 Flash (78,9 %). Pagal \u0161\u012f etalon\u0105 GPT-4o surinko 88,7%, o \u201eGemini Ultra\u201c gali pasigirti auk\u0161\u010diausiu balu \u2013 90% (vis d\u0117lto tai n\u0117ra ma\u017ei AI modeliai).<\/p>\n

\"\"
\u0160altinis: OpenAI<\/figcaption><\/figure>\n

GPT-4o mini \u0161iandien pristatomas \u201eChatGPT Free\u201c, \u201eTeam\u201c ir \u201ePlus\u201c naudotojams bei k\u016br\u0117jams. \u201eChatGPT\u201c naudotojams jis i\u0161 esm\u0117s pakeit\u0117 GPT-3.5; GPT-4o mini bus modelis, kuris bus naudojamas pokalbiui pagal numatytuosius nustatymus, kai pasibaigs nemokam\u0173 GPT-4o u\u017eklaus\u0173. K\u016br\u0117jai vis tiek tur\u0117s galimyb\u0119 naudoti GPT-3.5 per API, bet galiausiai jos bus atsisakyta. \u201eChatGPT Enterprise\u201c vartotojai kit\u0105 savait\u0119 gaus prieig\u0105 prie GPT-4o mini.<\/p>\n

Kaip min\u0117ta pirmiau, GPT-4o mini tikslas yra pad\u0117ti k\u016br\u0117jams rasti pig\u0173 ir delsos model\u012f savo programai, kuris taip pat b\u016bt\u0173 tinkamas. Palyginti su kitais ma\u017eais modeliais, GPT-4o mini puikiai atlieka teksto ir vizijos samprotavimo u\u017eduotis, matematines samprotavimo ir kodavimo u\u017eduotis, taip pat multimodalin\u012f samprotavim\u0105.<\/p>\n

\u0160iuo metu jis palaiko ir tekst\u0105, ir vizij\u0105 API, palaikydamas teksto, vaizdo, vaizdo ir garso \u012fvestis ir i\u0161vestis ateities plane.<\/p>\n

GPT-4o mini kontekstinis langas yra 128 000 \u012fvesties atpa\u017einimo ir 16 000 i\u0161vesties prieigos rakt\u0173 vienai u\u017eklausai, o jo \u017einios t\u0119siasi iki 2023 m. spalio m\u0117n. Jis taip pat gali gana ekonomi\u0161kai tvarkyti tekst\u0105 ne angl\u0173 kalba.<\/p>\n","protected":false},"excerpt":{"rendered":"

Kitas AI modelis prisijung\u0117 prie AI modeli\u0173 j\u016bros, o \u0161is v\u0117l yra i\u0161 OpenAI. OpenAI anks\u010diau \u0161iais metais i\u0161leido GPT-4o, kuris tur\u0117jo b\u016bti efektyvesnis nei GPT-4. Ta\u010diau tai vis tiek buvo gana brangu ir gal\u0117jo pareikalauti nema\u017eai s\u0105skait\u0173, ypa\u010d k\u016br\u0117jams, kuriems reikia pakartotinai per dien\u0105 i\u0161kviesti AI model\u012f per API savo programoms. Tod\u0117l k\u016br\u0117jai \u0117m\u0117 […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[176,177,123],"class_list":["post-2782","post","type-post","status-publish","format-standard","hentry","category-how-to","tag-artificial-intelligence","tag-chatgpt","tag-microsoft"],"acf":[],"_links":{"self":[{"href":"https:\/\/howtogeek.blog\/lt\/wp-json\/wp\/v2\/posts\/2782","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/howtogeek.blog\/lt\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/howtogeek.blog\/lt\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/howtogeek.blog\/lt\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/howtogeek.blog\/lt\/wp-json\/wp\/v2\/comments?post=2782"}],"version-history":[{"count":1,"href":"https:\/\/howtogeek.blog\/lt\/wp-json\/wp\/v2\/posts\/2782\/revisions"}],"predecessor-version":[{"id":2783,"href":"https:\/\/howtogeek.blog\/lt\/wp-json\/wp\/v2\/posts\/2782\/revisions\/2783"}],"wp:attachment":[{"href":"https:\/\/howtogeek.blog\/lt\/wp-json\/wp\/v2\/media?parent=2782"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/howtogeek.blog\/lt\/wp-json\/wp\/v2\/categories?post=2782"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/howtogeek.blog\/lt\/wp-json\/wp\/v2\/tags?post=2782"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}