{"id":167258,"date":"2026-04-03T05:11:55","date_gmt":"2026-04-03T03:11:56","guid":{"rendered":"https:\/\/prohoster.info\/blog\/novosti-interneta\/google-vypustil-otkrytuyu-ai-model-gemma-4-postroennuyu-na-tehnologiyah-gemini-3"},"modified":"2026-04-03T05:11:55","modified_gmt":"2026-04-03T03:11:56","slug":"google-vypustil-otkrytuyu-ai-model-gemma-4-postroennuyu-na-tehnologiyah-gemini-3","status":"publish","type":"post","link":"https:\/\/prohoster.info\/pl\/blog\/news\/google-vypustil-otkrytuyu-ai-model-gemma-4-postroennuyu-na-tehnologiyah-gemini-3","title":{"rendered":"Google wyda\u0142 otwarty model AI Gemma 4, zbudowany na technologiach Gemini 3.","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"

Firma Google opublikowa\u0142a now\u0105 rodzin\u0119 du\u017cych modeli j\u0119zykowych Gemma 4, opartych na technologii modelu Gemini 3. Gemma 4 jest dost\u0119pna na licencji Apache w wariantach z 2,3, 4,5, 25,2 i 30,7 miliardami parametr\u00f3w (E2B, E4B, 31B i 26B A4B). Warianty E2B i E4B nadaj\u0105 si\u0119 do u\u017cycia na urz\u0105dzeniach mobilnych, systemach Internetu Rzeczy (IoT) oraz p\u0142ytach typu Raspberry Pi, podczas gdy pozosta\u0142e warianty nadaj\u0105 si\u0119 do zastosowania na stacjach roboczych i systemach z konsumenckimi GPU. Rozmiar kontekstu rozwa\u017canego przez model wynosi 128 tysi\u0119cy token\u00f3w dla modeli E2B i E4B oraz 256 tysi\u0119cy token\u00f3w dla modeli 31B i 26B A4B. <\/p>\n

Modele s\u0105 wieloj\u0119zyczne i multimodalne: w zestawie wspieraj\u0105 35 j\u0119zyk\u00f3w (w trakcie szkolenia u\u017cyto ponad 140 j\u0119zyk\u00f3w), a na wej\u015bciu mog\u0105 przetwarza\u0107 tekst i obrazy (modele E2B i E4B dodatkowo wspieraj\u0105 przetwarzanie d\u017awi\u0119ku). Model 26B A4B oparty jest na architekturze MoE (Mixture-of-Experts), w kt\u00f3rej model jest podzielony na seri\u0119 sieci eksperckich (przy generowaniu odpowiedzi mog\u0105 by\u0107 u\u017cywane tylko 3,8 miliarda parametr\u00f3w, ale szybko\u015b\u0107 jest znacznie wy\u017csza ni\u017c w przypadku klasycznych du\u017cych modeli), podczas gdy pozosta\u0142e warianty wykorzystuj\u0105 klasyczn\u0105 monolityczn\u0105 architektur\u0119. <\/p>\n

Modele wspieraj\u0105 rozumowanie i konfigurowalne tryby my\u015blenia, obs\u0142uguj\u0105 rol\u0119 systemow\u0105 (System Role) do przetwarzania instrukcji (regu\u0142, ogranicze\u0144) oddzielnie od danych. Modele mog\u0105 by\u0107 u\u017cywane do pisania kodu, rozpoznawania obiekt\u00f3w na obrazach, analizy wideo klatka po klatce, analizy dokument\u00f3w i PDF, optycznego rozpoznawania tekstu (OCR), rozpoznawania mowy oraz t\u0142umaczenia mi\u0119dzy j\u0119zykami. Mo\u017cliwe jest u\u017cycie jako autonomicznych agent\u00f3w, kt\u00f3rzy wsp\u00f3\u0142dzia\u0142aj\u0105 z r\u00f3\u017cnymi narz\u0119dziami i API. <\/p>\n

W wi\u0119kszo\u015bci test\u00f3w modele serii Gemma 4 znacz\u0105co przewy\u017cszy\u0142y model Gemma 3 z 27 miliardami parametr\u00f3w. Wspierane jest u\u017cycie Gemma 4 z narz\u0119dziami i bibliotekami LiteRT-LM, vLLM, llama.cpp, MLX, Ollama, NVIDIA NIM i NeMo, LM Studio, Unsloth, SGLang, Cactus, Basetan, MaxText, Tunix i Keras.

\"Google<\/center>
\"Google<\/center>
\n
\u0179r\u00f3d\u0142o: opennet.ru<\/a> <\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"

\u041a\u043e\u043c\u043f\u0430\u043d\u0438\u044f Google \u043e\u043f\u0443\u0431\u043b\u0438\u043a\u043e\u0432\u0430\u043b\u0430 \u043d\u043e\u0432\u043e\u0435 \u0441\u0435\u043c\u0435\u0439\u0441\u0442\u0432\u043e \u0431\u043e\u043b\u044c\u0448\u0438\u0445 \u044f\u0437\u044b\u043a\u043e\u0432\u044b\u0445 \u043c\u043e\u0434\u0435\u043b\u0435\u0439 Gemma 4, \u043e\u0441\u043d\u043e\u0432\u0430\u043d\u043d\u044b\u0445 \u043d\u0430 \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0438\u044f\u0445 \u043c\u043e\u0434\u0435\u043b\u0438 Gemini 3. Gemma 4 \u0440\u0430\u0441\u043f\u0440\u043e\u0441\u0442\u0440\u0430\u043d\u044f\u0435\u0442\u0441\u044f \u043f\u043e\u0434 \u043b\u0438\u0446\u0435\u043d\u0437\u0438\u0435\u0439 Apache \u0432 \u0432\u0430\u0440\u0438\u0430\u043d\u0442\u0430\u0445 \u0441 2.3, 4.5, 25.2 \u0438 30.7 \u043c\u0438\u043b\u043b\u0438\u0430\u0440\u0434\u0430\u043c\u0438 \u043f\u0430\u0440\u0430\u043c\u0435\u0442\u0440\u043e\u0432 (E2B, E4B, 31B \u0438 26B A4B). \u0412\u0430\u0440\u0438\u0430\u043d\u0442\u044b E2B \u0438 E4B \u043f\u043e\u0434\u0445\u043e\u0434\u044f\u0442 \u0434\u043b\u044f \u0438\u0441\u043f\u043e\u043b\u044c\u0437\u043e\u0432\u0430\u043d\u0438\u044f \u043d\u0430 \u043c\u043e\u0431\u0438\u043b\u044c\u043d\u044b\u0445 \u0443\u0441\u0442\u0440\u043e\u0439\u0441\u0442\u0432\u0430\u0445, \u0441\u0438\u0441\u0442\u0435\u043c\u0430\u0445 \u0438\u043d\u0442\u0435\u0440\u043d\u0435\u0442\u0430 \u0432\u0435\u0449\u0435\u0439 (IoT) \u0438 \u043f\u043b\u0430\u0442\u0430\u0445 \u0442\u0438\u043f\u0430 […]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":8,"featured_media":167259,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[702],"tags":[],"class_list":["post-167258","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-news"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\n","aioseo_head_json":{"title":"\ud83e\udd47Google wyda\u0142 otwarty model AI Gemma 4, oparty na technologii Gemini 3 | ProHoster","description":"Firma Google opublikowa\u0142a now\u0105 rodzin\u0119 du\u017cych modeli j\u0119zykowych Gemma 4, opartych na technologii modelu Gemini 3.","canonical_url":"https:\/\/prohoster.info\/pl\/blog\/news\/google-vypustil-otkrytuyu-ai-model-gemma-4-postroennuyu-na-tehnologiyah-gemini-3","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"pl_PL","og:site_name":"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b","og:type":"article","og:title":"\ud83e\udd47Google \u0432\u044b\u043f\u0443\u0441\u0442\u0438\u043b \u043e\u0442\u043a\u0440\u044b\u0442\u0443\u044e AI-\u043c\u043e\u0434\u0435\u043b\u044c Gemma 4, \u043f\u043e\u0441\u0442\u0440\u043e\u0435\u043d\u043d\u0443\u044e \u043d\u0430 \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0438\u044f\u0445 Gemini 3 | ProHoster","og:description":"\u041a\u043e\u043c\u043f\u0430\u043d\u0438\u044f Google \u043e\u043f\u0443\u0431\u043b\u0438\u043a\u043e\u0432\u0430\u043b\u0430 \u043d\u043e\u0432\u043e\u0435 \u0441\u0435\u043c\u0435\u0439\u0441\u0442\u0432\u043e \u0431\u043e\u043b\u044c\u0448\u0438\u0445 \u044f\u0437\u044b\u043a\u043e\u0432\u044b\u0445 \u043c\u043e\u0434\u0435\u043b\u0435\u0439 Gemma 4, \u043e\u0441\u043d\u043e\u0432\u0430\u043d\u043d\u044b\u0445 \u043d\u0430 \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0438\u044f\u0445 \u043c\u043e\u0434\u0435\u043b\u0438 Gemini 3.","og:url":"https:\/\/prohoster.info\/pl\/blog\/news\/google-vypustil-otkrytuyu-ai-model-gemma-4-postroennuyu-na-tehnologiyah-gemini-3","og:image":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:secure_url":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:width":350,"og:image:height":350,"article:published_time":"2026-04-03T03:11:56+00:00","article:modified_time":"2026-04-03T03:11:56+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":[],"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/pl\/wp-json\/wp\/v2\/posts\/167258","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/pl\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/pl\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/pl\/wp-json\/wp\/v2\/users\/8"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/pl\/wp-json\/wp\/v2\/comments?post=167258"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/pl\/wp-json\/wp\/v2\/posts\/167258\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/prohoster.info\/pl\/wp-json\/wp\/v2\/media\/167259"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/pl\/wp-json\/wp\/v2\/media?parent=167258"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/pl\/wp-json\/wp\/v2\/categories?post=167258"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/pl\/wp-json\/wp\/v2\/tags?post=167258"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}