{"id":12933,"date":"2024-08-04T16:42:03","date_gmt":"2024-08-04T21:42:03","guid":{"rendered":"http:\/\/skimai.com\/?p=12933"},"modified":"2024-08-21T18:22:21","modified_gmt":"2024-08-21T23:22:21","slug":"como-o-metas-llama-3-1-esta-a-alargar-os-limites-da-ia-de-fonte-aberta","status":"publish","type":"post","link":"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/","title":{"rendered":"Meta's Llama 3.1: Ultrapassar os limites da IA de c\u00f3digo aberto"},"content":{"rendered":"<p>A Meta anunciou recentemente <a rel=\"noopener noreferrer\" href=\"https:\/\/llama.meta.com\/\">Lhama 3.1<\/a>O software de intelig\u00eancia artificial (IA) da Microsoft, o seu modelo de grande linguagem (LLM) de c\u00f3digo aberto mais avan\u00e7ado at\u00e9 \u00e0 data. Este lan\u00e7amento representa um marco significativo na democratiza\u00e7\u00e3o da tecnologia de IA, potencialmente colmatando o fosso entre modelos de c\u00f3digo aberto e modelos propriet\u00e1rios.<\/p>\n\n\n<p>O Llama 3.1 \u00e9 um grande salto em frente nas capacidades de IA de c\u00f3digo aberto. Com o seu modelo emblem\u00e1tico de 405 mil milh\u00f5es de par\u00e2metros, a Meta est\u00e1 a desafiar a no\u00e7\u00e3o de que a IA de ponta tem de ser de c\u00f3digo fechado e propriet\u00e1ria. Este lan\u00e7amento assinala uma nova era em que as capacidades de IA de ponta est\u00e3o acess\u00edveis a investigadores, programadores e empresas de todas as dimens\u00f5es.<\/p>\n\n\n<p>As principais melhorias na Llama 3.1 incluem um comprimento de contexto expandido de 128.000 tokens, suporte para oito idiomas e desempenho incompar\u00e1vel em \u00e1reas como racioc\u00ednio, matem\u00e1tica e gera\u00e7\u00e3o de c\u00f3digo. Esses avan\u00e7os posicionam a Llama 3.1 como uma ferramenta vers\u00e1til capaz de lidar com tarefas complexas do mundo real em v\u00e1rios dom\u00ednios no ambiente corporativo.<\/p>\n\n\n<div id=\"ez-toc-container\" class=\"ez-toc-v2_0_82_1 counter-hierarchy ez-toc-counter ez-toc-grey ez-toc-container-direction\">\n<div class=\"ez-toc-title-container\">\n<p class=\"ez-toc-title\" style=\"cursor:inherit\">\u00cdndice<\/p>\n<span class=\"ez-toc-title-toggle\"><a href=\"#\" class=\"ez-toc-pull-right ez-toc-btn ez-toc-btn-xs ez-toc-btn-default ez-toc-toggle\" aria-label=\"Alternar o \u00edndice\"><span class=\"ez-toc-js-icon-con\"><span class=\"\"><span class=\"eztoc-hide\" style=\"display:none;\">Alternar<\/span><span class=\"ez-toc-icon-toggle-span\"><svg style=\"fill: #999;color:#999\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" class=\"list-377408\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" fill=\"none\"><path d=\"M6 6H4v2h2V6zm14 0H8v2h12V6zM4 11h2v2H4v-2zm16 0H8v2h12v-2zM4 16h2v2H4v-2zm16 0H8v2h12v-2z\" fill=\"currentColor\"><\/path><\/svg><svg style=\"fill: #999;color:#999\" class=\"arrow-unsorted-368013\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"10px\" height=\"10px\" viewbox=\"0 0 24 24\" version=\"1.2\" baseprofile=\"tiny\"><path d=\"M18.2 9.3l-6.2-6.3-6.2 6.3c-.2.2-.3.4-.3.7s.1.5.3.7c.2.2.4.3.7.3h11c.3 0 .5-.1.7-.3.2-.2.3-.5.3-.7s-.1-.5-.3-.7zM5.8 14.7l6.2 6.3 6.2-6.3c.2-.2.3-.5.3-.7s-.1-.5-.3-.7c-.2-.2-.4-.3-.7-.3h-11c-.3 0-.5.1-.7.3-.2.2-.3.5-.3.7s.1.5.3.7z\"\/><\/svg><\/span><\/span><\/span><\/a><\/span><\/div>\n<nav><ul class='ez-toc-list ez-toc-list-level-1' ><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-1\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#The_Evolution_of_Llama_From_2_to_31\" >A evolu\u00e7\u00e3o do Llama: do 2 ao 3.1<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-2\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#Technical_Specifications_of_Llama_31\" >Especifica\u00e7\u00f5es t\u00e9cnicas do Llama 3.1<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-3\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#Breakthrough_Capabilities\" >Capacidades inovadoras<\/a><ul class='ez-toc-list-level-3' ><li class='ez-toc-heading-level-3'><a class=\"ez-toc-link ez-toc-heading-4\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#Expanded_Context_Length\" >Comprimento do contexto expandido<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-3'><a class=\"ez-toc-link ez-toc-heading-5\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#Multilingual_Support\" >Suporte multilingue<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-3'><a class=\"ez-toc-link ez-toc-heading-6\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#Advanced_Reasoning_and_Tool_Use\" >Racioc\u00ednio avan\u00e7ado e utiliza\u00e7\u00e3o de ferramentas<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-3'><a class=\"ez-toc-link ez-toc-heading-7\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#Code_Generation_and_Math_Prowess\" >Gera\u00e7\u00e3o de c\u00f3digo e habilidade matem\u00e1tica<\/a><\/li><\/ul><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-8\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#The_Open-Source_Advantage\" >A vantagem do c\u00f3digo aberto<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-9\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#Ecosystem_and_Deployment\" >Ecossistema e implanta\u00e7\u00e3o<\/a><ul class='ez-toc-list-level-3' ><li class='ez-toc-heading-level-3'><a class=\"ez-toc-link ez-toc-heading-10\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#Partner_Integrations\" >Integra\u00e7\u00f5es de parceiros<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-3'><a class=\"ez-toc-link ez-toc-heading-11\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#Inference_Optimization_and_Scalability\" >Otimiza\u00e7\u00e3o e escalabilidade da infer\u00eancia<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-3'><a class=\"ez-toc-link ez-toc-heading-12\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#The_Llama_Stack_and_Standardization_Efforts\" >A pilha Llama e os esfor\u00e7os de normaliza\u00e7\u00e3o<\/a><\/li><\/ul><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-13\" href=\"https:\/\/skimai.com\/pt\/how-metas-llama-3-1-is-pushing-the-boundaries-of-open-source-ai\/#Llama_31s_Promise_and_Potential\" >Promessas e potencialidades da Llama 3.1<\/a><\/li><\/ul><\/nav><\/div>\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"The_Evolution_of_Llama_From_2_to_31\"><\/span>A evolu\u00e7\u00e3o do Llama: do 2 ao 3.1<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n<p>Para apreciar o significado do Llama 3.1, vale a pena revisitar os seus antecessores. O Llama 2, lan\u00e7ado em 2023, j\u00e1 era um grande passo em frente na IA de c\u00f3digo aberto. Oferecia modelos que variavam entre 7B e 70B par\u00e2metros e demonstrava um desempenho competitivo em v\u00e1rios benchmarks.<\/p>\n\n\n<p><strong><u>A Llama 3.1 assenta nesta base com v\u00e1rios avan\u00e7os importantes:<\/u><\/strong><\/p>\n\n\n<ol class=\"wp-block-list\">\n<li><p><strong>Aumento do tamanho do modelo:<\/strong> A introdu\u00e7\u00e3o do modelo de par\u00e2metros 405B alarga os limites do que \u00e9 poss\u00edvel na IA de c\u00f3digo aberto.<\/p><\/li><li><p><strong>Comprimento do contexto alargado:<\/strong> De 4K tokens no Llama 2 para 128K no Llama 3.1, permitindo uma compreens\u00e3o mais complexa e matizada de textos mais longos.<\/p><\/li><li><p><strong>Capacidades multilingues:<\/strong> O suporte lingu\u00edstico alargado permite aplica\u00e7\u00f5es mais diversificadas em diferentes regi\u00f5es e casos de utiliza\u00e7\u00e3o.<\/p><\/li><li><p><strong>Melhoria do racioc\u00ednio e das tarefas especializadas:<\/strong> Desempenho melhorado em \u00e1reas como o racioc\u00ednio matem\u00e1tico e a gera\u00e7\u00e3o de c\u00f3digo.<\/p><\/li>\n<\/ol>\n\n\n<p>Quando comparado com modelos de c\u00f3digo fechado, como o GPT-4 e o Claude 3.5 Sonnet, o Llama 3.1 405B mant\u00e9m-se firme em v\u00e1rios benchmarks. Este n\u00edvel de desempenho num modelo de c\u00f3digo aberto n\u00e3o tem precedentes.<\/p>\n\n\n<figure class=\"wp-block-image\">\n<img decoding=\"async\" src=\"http:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/ceb48833-8f7d-4551-a3a2-8e0936a5105e.png\" alt=\"Benchmarks do Meta Llama 3.1\" \/>\n<\/figure>\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Technical_Specifications_of_Llama_31\"><\/span>Especifica\u00e7\u00f5es t\u00e9cnicas do Llama 3.1<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n<p>Quanto aos pormenores t\u00e9cnicos, o Llama 3.1 oferece uma gama de tamanhos de modelos para se adaptar a diferentes necessidades e recursos computacionais:<\/p>\n\n\n<ol class=\"wp-block-list\">\n<li><p><strong>Modelo de par\u00e2metros 8B: <\/strong>Adequado para aplica\u00e7\u00f5es ligeiras e dispositivos de ponta.<\/p><\/li><li><p><strong>Modelo de par\u00e2metros 70B:<\/strong> Um equil\u00edbrio entre os requisitos de desempenho e de recursos.<\/p><\/li><li><p><strong>Modelo de par\u00e2metros 405B:<\/strong> O modelo emblem\u00e1tico, que ultrapassa os limites das capacidades de IA de fonte aberta.<\/p><\/li>\n<\/ol>\n\n\n<p>A metodologia de treino para o Llama 3.1 envolveu um conjunto de dados maci\u00e7o de mais de 15 trili\u00f5es de tokens, significativamente maior do que os seus antecessores. Estes dados de treino extensivos, combinados com t\u00e9cnicas refinadas de curadoria e pr\u00e9-processamento de dados, contribuem para o melhor desempenho e versatilidade do modelo.<\/p>\n\n\n<p>Arquitetonicamente, a Llama 3.1 mant\u00e9m um modelo de transformador apenas de descodificador, dando prioridade \u00e0 estabilidade do treino em detrimento de abordagens mais experimentais como a mistura de especialistas. No entanto, o Meta implementou v\u00e1rias optimiza\u00e7\u00f5es para permitir uma forma\u00e7\u00e3o e infer\u00eancia eficientes a esta escala sem precedentes:<\/p>\n\n\n<ol class=\"wp-block-list\">\n<li><p><strong>Infraestrutura de forma\u00e7\u00e3o escal\u00e1vel: <\/strong>Utilizando mais de 16.000 GPUs H100 para treinar o modelo 405B.<\/p><\/li><li><p><strong>Procedimento iterativo de p\u00f3s-treino: <\/strong>Utilizar a afina\u00e7\u00e3o supervisionada e a otimiza\u00e7\u00e3o direta das prefer\u00eancias para melhorar capacidades espec\u00edficas.<\/p><\/li><li><p><strong>T\u00e9cnicas de quantiza\u00e7\u00e3o: <\/strong>Reduzir o modelo de 16 bits para 8 bits num\u00e9ricos para uma infer\u00eancia mais eficiente, permitindo a implementa\u00e7\u00e3o em n\u00f3s de servidor \u00fanico.<\/p><\/li>\n<\/ol>\n\n\n<p>Estas escolhas t\u00e9cnicas reflectem um equil\u00edbrio entre ultrapassar os limites da dimens\u00e3o do modelo e garantir uma utiliza\u00e7\u00e3o pr\u00e1tica numa s\u00e9rie de cen\u00e1rios de implanta\u00e7\u00e3o.<\/p>\n\n\n<p>Ao disponibilizar abertamente estes modelos avan\u00e7ados, a Meta n\u00e3o est\u00e1 apenas a partilhar um produto, mas a fornecer uma plataforma para a inova\u00e7\u00e3o. As especifica\u00e7\u00f5es t\u00e9cnicas do Llama 3.1 abrem novas possibilidades para os investigadores e programadores explorarem aplica\u00e7\u00f5es de IA de ponta, acelerando o ritmo do avan\u00e7o da IA em toda a ind\u00fastria.<\/p>\n\n\n<figure class=\"wp-block-image\">\n<img decoding=\"async\" src=\"http:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/5abb38bb-3e47-4c4e-a442-b87a4867748a.png\" alt=\"Arquitetura do Meta Llama 3.1\" \/>\n<\/figure>\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Breakthrough_Capabilities\"><\/span>Capacidades inovadoras<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n<p>O Llama 3.1 apresenta v\u00e1rias capacidades inovadoras que o distinguem no panorama da IA:<\/p>\n\n\n<h3 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Expanded_Context_Length\"><\/span>Comprimento do contexto expandido<span class=\"ez-toc-section-end\"><\/span><\/h3>\n\n\n<p>O salto para uma janela de contexto de token de 128K \u00e9 um divisor de \u00e1guas. Esta capacidade alargada permite \u00e0 Llama 3.1 processar e compreender textos muito mais longos, permitindo:<\/p>\n\n\n<ul class=\"wp-block-list\">\n<li><p>An\u00e1lise exaustiva de documentos<\/p><\/li><li><p>Gera\u00e7\u00e3o de conte\u00fados de formato longo<\/p><\/li><li><p>Tratamento de conversas com mais nuances<\/p><\/li>\n<\/ul>\n\n\n<p>Esta funcionalidade abre novas possibilidades para aplica\u00e7\u00f5es em \u00e1reas como o processamento de documentos jur\u00eddicos, a an\u00e1lise de literatura e a resolu\u00e7\u00e3o de problemas complexos que exijam a reten\u00e7\u00e3o e a s\u00edntese de grandes quantidades de informa\u00e7\u00e3o.<\/p>\n\n\n<h3 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Multilingual_Support\"><\/span>Suporte multilingue<span class=\"ez-toc-section-end\"><\/span><\/h3>\n\n\n<p>O suporte da Llama 3.1 para oito l\u00ednguas alarga significativamente a sua aplicabilidade global. Esta capacidade multilingue:<\/p>\n\n\n<ul class=\"wp-block-list\">\n<li><p>Melhora a comunica\u00e7\u00e3o intercultural<\/p><\/li><li><p>Permite aplica\u00e7\u00f5es de IA mais inclusivas<\/p><\/li><li><p>Apoia as opera\u00e7\u00f5es comerciais globais<\/p><\/li>\n<\/ul>\n\n\n<p>Ao quebrar as barreiras lingu\u00edsticas, o Llama 3.1 abre caminho a solu\u00e7\u00f5es de IA mais diversificadas e orientadas para o mundo.<\/p>\n\n\n<h3 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Advanced_Reasoning_and_Tool_Use\"><\/span>Racioc\u00ednio avan\u00e7ado e utiliza\u00e7\u00e3o de ferramentas<span class=\"ez-toc-section-end\"><\/span><\/h3>\n\n\n<p>O modelo demonstra capacidades de racioc\u00ednio sofisticadas e a capacidade de utilizar eficazmente ferramentas externas. Este avan\u00e7o manifesta-se em:<\/p>\n\n\n<ul class=\"wp-block-list\">\n<li><p>Melhoria da dedu\u00e7\u00e3o l\u00f3gica e da resolu\u00e7\u00e3o de problemas<\/p><\/li><li><p>Capacidade acrescida de seguir instru\u00e7\u00f5es complexas<\/p><\/li><li><p>Utiliza\u00e7\u00e3o eficaz de bases de conhecimento externas e APIs<\/p><\/li>\n<\/ul>\n\n\n<p>Estas capacidades fazem do Llama 3.1 uma ferramenta poderosa para tarefas que exigem compet\u00eancias cognitivas de alto n\u00edvel, desde o planeamento estrat\u00e9gico \u00e0 an\u00e1lise de dados complexos.<\/p>\n\n\n<h3 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Code_Generation_and_Math_Prowess\"><\/span>Gera\u00e7\u00e3o de c\u00f3digo e habilidade matem\u00e1tica<span class=\"ez-toc-section-end\"><\/span><\/h3>\n\n\n<p>O Llama 3.1 demonstra capacidades not\u00e1veis nos dom\u00ednios t\u00e9cnicos:<\/p>\n\n\n<ul class=\"wp-block-list\">\n<li><p>Gera\u00e7\u00e3o de c\u00f3digo funcional de alta qualidade em v\u00e1rias linguagens de programa\u00e7\u00e3o<\/p><\/li><li><p>Resolver problemas matem\u00e1ticos complexos com exatid\u00e3o<\/p><\/li><li><p>Assist\u00eancia na conce\u00e7\u00e3o e otimiza\u00e7\u00e3o de algoritmos<\/p><\/li>\n<\/ul>\n\n\n<p>Estas caracter\u00edsticas posicionam o Llama 3.1 como um ativo valioso para o desenvolvimento de software, computa\u00e7\u00e3o cient\u00edfica e aplica\u00e7\u00f5es de engenharia.<\/p>\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"The_Open-Source_Advantage\"><\/span>A vantagem do c\u00f3digo aberto<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n<p>A natureza de c\u00f3digo aberto da Llama 3.1 traz v\u00e1rios benef\u00edcios significativos.<\/p>\n\n\n<p>Ao disponibilizar gratuitamente capacidades de IA de ponta, a Meta est\u00e1 a faz\u00ea-lo:<\/p>\n\n\n<ul class=\"wp-block-list\">\n<li><p>Reduzir os obst\u00e1culos \u00e0 entrada na investiga\u00e7\u00e3o e no desenvolvimento da IA<\/p><\/li><li><p>Permitir que as organiza\u00e7\u00f5es mais pequenas e os programadores individuais tirem partido da IA avan\u00e7ada<\/p><\/li><li><p>Promover um ecossistema de IA mais diversificado e inovador<\/p><\/li>\n<\/ul>\n\n\n<p>Esta democratiza\u00e7\u00e3o poder\u00e1 conduzir a uma prolifera\u00e7\u00e3o de aplica\u00e7\u00f5es de IA em v\u00e1rios sectores, acelerando potencialmente o progresso tecnol\u00f3gico.<\/p>\n\n\n<p>A capacidade de aceder e modificar os pesos dos modelos da Llama 3.1 abre oportunidades de personaliza\u00e7\u00e3o sem precedentes:<\/p>\n\n\n<ul class=\"wp-block-list\">\n<li><p>Adapta\u00e7\u00e3o a dom\u00ednios espec\u00edficos para ind\u00fastrias especializadas<\/p><\/li><li><p>Ajuste fino para casos de utiliza\u00e7\u00e3o e conjuntos de dados \u00fanicos<\/p><\/li><li><p>Experimenta\u00e7\u00e3o de novas t\u00e9cnicas e arquitecturas de forma\u00e7\u00e3o<\/p><\/li>\n<\/ul>\n\n\n<p>Esta flexibilidade permite que as organiza\u00e7\u00f5es adaptem o modelo \u00e0s suas necessidades espec\u00edficas, conduzindo potencialmente a solu\u00e7\u00f5es de IA mais eficazes e eficientes.<\/p>\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Ecosystem_and_Deployment\"><\/span>Ecossistema e implanta\u00e7\u00e3o<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n<p>O lan\u00e7amento da Llama 3.1 \u00e9 acompanhado por um ecossistema robusto para apoiar a sua implanta\u00e7\u00e3o e utiliza\u00e7\u00e3o:<\/p>\n\n\n<h3 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Partner_Integrations\"><\/span>Integra\u00e7\u00f5es de parceiros<span class=\"ez-toc-section-end\"><\/span><\/h3>\n\n\n<p>A Meta colaborou com os l\u00edderes do sector para garantir um apoio alargado \u00e0 Llama 3.1:<\/p>\n\n\n<ul class=\"wp-block-list\">\n<li><p>Os fornecedores de servi\u00e7os na nuvem, como o AWS, o Google Cloud e o Azure, oferecem op\u00e7\u00f5es de implementa\u00e7\u00e3o sem descontinuidades<\/p><\/li><li><p>Os fabricantes de hardware, como a NVIDIA e a Dell, fornecem uma infraestrutura optimizada<\/p><\/li><li><p>Plataformas de dados como Databricks e Snowflake permitem o processamento eficiente de dados e a integra\u00e7\u00e3o de modelos<\/p><\/li>\n<\/ul>\n\n\n<p>Essas parcerias garantem que as organiza\u00e7\u00f5es possam aproveitar a Llama 3.1 em seus conjuntos de tecnologia existentes.<\/p>\n\n\n<figure class=\"wp-block-image\">\n<img decoding=\"async\" src=\"http:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/f65464b4-ea93-4bca-ae3d-c4f0bb54d5a3.png\" alt=\"Caracter\u00edsticas do Meta Llama 3.1\" \/>\n<\/figure>\n\n\n<h3 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Inference_Optimization_and_Scalability\"><\/span>Otimiza\u00e7\u00e3o e escalabilidade da infer\u00eancia<span class=\"ez-toc-section-end\"><\/span><\/h3>\n\n\n<p>Para tornar a Llama 3.1 pr\u00e1tica para as aplica\u00e7\u00f5es do mundo real, foram implementadas v\u00e1rias optimiza\u00e7\u00f5es:<\/p>\n\n\n<ul class=\"wp-block-list\">\n<li><p>As t\u00e9cnicas de quantiza\u00e7\u00e3o reduzem os requisitos computacionais do modelo<\/p><\/li><li><p>Motores de infer\u00eancia optimizados como o vLLM e o TensorRT aumentam o desempenho<\/p><\/li><li><p>As op\u00e7\u00f5es de implementa\u00e7\u00e3o escal\u00e1veis respondem a v\u00e1rios casos de utiliza\u00e7\u00e3o, desde dispositivos perif\u00e9ricos a centros de dados<\/p><\/li>\n<\/ul>\n\n\n<p>Estas optimiza\u00e7\u00f5es tornam vi\u00e1vel a implementa\u00e7\u00e3o at\u00e9 do modelo de par\u00e2metros 405B em ambientes de produ\u00e7\u00e3o.<\/p>\n\n\n<h3 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"The_Llama_Stack_and_Standardization_Efforts\"><\/span>A pilha Llama e os esfor\u00e7os de normaliza\u00e7\u00e3o<span class=\"ez-toc-section-end\"><\/span><\/h3>\n\n\n<p>A Meta est\u00e1 a promover a normaliza\u00e7\u00e3o no ecossistema de IA:<\/p>\n\n\n<ul class=\"wp-block-list\">\n<li><p>A Llama Stack proposta tem por objetivo criar uma interface comum para os componentes de IA<\/p><\/li><li><p>As API normalizadas poder\u00e3o facilitar a integra\u00e7\u00e3o e a interoperabilidade entre diferentes ferramentas e plataformas de IA<\/p><\/li><li><p>Esta iniciativa poder\u00e1 conduzir a um ecossistema de desenvolvimento de IA mais coeso e eficiente<\/p><\/li>\n<\/ul>\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Llama_31s_Promise_and_Potential\"><\/span>Promessas e potencialidades da Llama 3.1<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n<p>O lan\u00e7amento do Llama 3.1 pela Meta marca um momento crucial no panorama da IA, democratizando o acesso a capacidades de IA de n\u00edvel avan\u00e7ado. Ao oferecer um modelo de par\u00e2metros 405B com desempenho de ponta, suporte multilingue e comprimento de contexto alargado, tudo numa estrutura de c\u00f3digo aberto, a Meta estabeleceu um novo padr\u00e3o para uma IA acess\u00edvel e poderosa. Este passo n\u00e3o s\u00f3 desafia o dom\u00ednio dos modelos de c\u00f3digo fechado, como tamb\u00e9m abre caminho a uma inova\u00e7\u00e3o e colabora\u00e7\u00e3o sem precedentes na comunidade de IA. <\/p>\n\n\n<p>Ao nos encontrarmos nesta encruzilhada do desenvolvimento de IA, a Llama 3.1 representa mais do que apenas um avan\u00e7o tecnol\u00f3gico; ela incorpora uma vis\u00e3o de um futuro mais aberto, inclusivo e din\u00e2mico para a intelig\u00eancia artificial. O verdadeiro impacto desta vers\u00e3o ser\u00e1 revelado \u00e0 medida que os programadores, investigadores e empresas de todo o mundo aproveitarem o seu potencial, remodelando as ind\u00fastrias e ultrapassando os limites do que \u00e9 poss\u00edvel fazer com os LLM.<\/p>","protected":false},"excerpt":{"rendered":"<p>Meta has recently announced Llama 3.1, its most advanced open-source large language model (LLM) to date. This release marks a significant milestone in the democratization of AI technology, potentially bridging the gap between open-source and proprietary models. Llama 3.1 is a big leap forward in open-source AI capabilities. With its flagship 405 billion parameter model, [&hellip;]<\/p>\n","protected":false},"author":1003,"featured_media":12938,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"single-custom-post-template.php","format":"standard","meta":{"_et_pb_use_builder":"","_et_pb_old_content":"","_et_gb_content_width":"","footnotes":""},"categories":[125,100,67],"tags":[],"class_list":["post-12933","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-enterprise-ai-blog","category-generative-ai","category-ml-nlp"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v24.1 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Meta&#039;s Llama 3.1: Pushing the Boundaries of Open-Source AI - Skim AI<\/title>\n<meta name=\"description\" content=\"Explore Llama 3.1 by Meta, a groundbreaking open-source large language model with 405 billion parameters. Discover its advanced features and enterprise benefits.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/skimai.com\/pt\/como-o-metas-llama-3-1-esta-a-alargar-os-limites-da-ia-de-fonte-aberta\/\" \/>\n<meta property=\"og:locale\" content=\"pt_PT\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Meta&#039;s Llama 3.1: Pushing the Boundaries of Open-Source AI - Skim AI\" \/>\n<meta property=\"og:description\" content=\"Explore Llama 3.1 by Meta, a groundbreaking open-source large language model with 405 billion parameters. Discover its advanced features and enterprise benefits.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/skimai.com\/pt\/como-o-metas-llama-3-1-esta-a-alargar-os-limites-da-ia-de-fonte-aberta\/\" \/>\n<meta property=\"og:site_name\" content=\"Skim AI\" \/>\n<meta property=\"article:published_time\" content=\"2024-08-04T21:42:03+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2024-08-21T23:22:21+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/llama-comparison.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1092\" \/>\n\t<meta property=\"og:image:height\" content=\"612\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Greggory Elias\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Escrito por\" \/>\n\t<meta name=\"twitter:data1\" content=\"Greggory Elias\" \/>\n\t<meta name=\"twitter:label2\" content=\"Tempo estimado de leitura\" \/>\n\t<meta name=\"twitter:data2\" content=\"7 minutos\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/\"},\"author\":{\"name\":\"Greggory Elias\",\"@id\":\"https:\/\/skimai.com\/uk\/#\/schema\/person\/7a883b4a2d2ea22040f42a7975eb86c6\"},\"headline\":\"Meta&#8217;s Llama 3.1: Pushing the Boundaries of Open-Source AI\",\"datePublished\":\"2024-08-04T21:42:03+00:00\",\"dateModified\":\"2024-08-21T23:22:21+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/\"},\"wordCount\":1234,\"publisher\":{\"@id\":\"https:\/\/skimai.com\/uk\/#organization\"},\"image\":{\"@id\":\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/llama-comparison.png\",\"articleSection\":[\"Enterprise AI\",\"Generative AI\",\"LLMs \/ NLP\"],\"inLanguage\":\"pt-PT\"},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/\",\"url\":\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/\",\"name\":\"Meta's Llama 3.1: Pushing the Boundaries of Open-Source AI - Skim AI\",\"isPartOf\":{\"@id\":\"https:\/\/skimai.com\/uk\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#primaryimage\"},\"image\":{\"@id\":\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/llama-comparison.png\",\"datePublished\":\"2024-08-04T21:42:03+00:00\",\"dateModified\":\"2024-08-21T23:22:21+00:00\",\"description\":\"Explore Llama 3.1 by Meta, a groundbreaking open-source large language model with 405 billion parameters. Discover its advanced features and enterprise benefits.\",\"breadcrumb\":{\"@id\":\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#breadcrumb\"},\"inLanguage\":\"pt-PT\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"pt-PT\",\"@id\":\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#primaryimage\",\"url\":\"https:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/llama-comparison.png\",\"contentUrl\":\"https:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/llama-comparison.png\",\"width\":1092,\"height\":612,\"caption\":\"llama comparison\"},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/skimai.com\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"Meta&#8217;s Llama 3.1: Pushing the Boundaries of Open-Source AI\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/skimai.com\/uk\/#website\",\"url\":\"https:\/\/skimai.com\/uk\/\",\"name\":\"Skim AI\",\"description\":\"The AI Agent Workforce Platform\",\"publisher\":{\"@id\":\"https:\/\/skimai.com\/uk\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/skimai.com\/uk\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"pt-PT\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/skimai.com\/uk\/#organization\",\"name\":\"Skim AI\",\"url\":\"https:\/\/skimai.com\/uk\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"pt-PT\",\"@id\":\"https:\/\/skimai.com\/uk\/#\/schema\/logo\/image\/\",\"url\":\"http:\/\/skimai.com\/wp-content\/uploads\/2020\/07\/SKIM-AI-Header-Logo.png\",\"contentUrl\":\"http:\/\/skimai.com\/wp-content\/uploads\/2020\/07\/SKIM-AI-Header-Logo.png\",\"width\":194,\"height\":58,\"caption\":\"Skim AI\"},\"image\":{\"@id\":\"https:\/\/skimai.com\/uk\/#\/schema\/logo\/image\/\"},\"sameAs\":[\"https:\/\/www.linkedin.com\/company\/skim-ai\"]},{\"@type\":\"Person\",\"@id\":\"https:\/\/skimai.com\/uk\/#\/schema\/person\/7a883b4a2d2ea22040f42a7975eb86c6\",\"name\":\"Greggory Elias\",\"url\":\"https:\/\/skimai.com\/pt\/author\/gregg\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"Meta's Llama 3.1: Ultrapassar os limites da IA de c\u00f3digo aberto - Skim AI","description":"Explore o Llama 3.1 da Meta, um modelo lingu\u00edstico inovador de c\u00f3digo aberto com 405 mil milh\u00f5es de par\u00e2metros. Descubra as suas funcionalidades avan\u00e7adas e as vantagens para a empresa.","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/skimai.com\/pt\/como-o-metas-llama-3-1-esta-a-alargar-os-limites-da-ia-de-fonte-aberta\/","og_locale":"pt_PT","og_type":"article","og_title":"Meta's Llama 3.1: Pushing the Boundaries of Open-Source AI - Skim AI","og_description":"Explore Llama 3.1 by Meta, a groundbreaking open-source large language model with 405 billion parameters. Discover its advanced features and enterprise benefits.","og_url":"https:\/\/skimai.com\/pt\/como-o-metas-llama-3-1-esta-a-alargar-os-limites-da-ia-de-fonte-aberta\/","og_site_name":"Skim AI","article_published_time":"2024-08-04T21:42:03+00:00","article_modified_time":"2024-08-21T23:22:21+00:00","og_image":[{"width":1092,"height":612,"url":"https:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/llama-comparison.png","type":"image\/png"}],"author":"Greggory Elias","twitter_card":"summary_large_image","twitter_misc":{"Escrito por":"Greggory Elias","Tempo estimado de leitura":"7 minutos"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#article","isPartOf":{"@id":"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/"},"author":{"name":"Greggory Elias","@id":"https:\/\/skimai.com\/uk\/#\/schema\/person\/7a883b4a2d2ea22040f42a7975eb86c6"},"headline":"Meta&#8217;s Llama 3.1: Pushing the Boundaries of Open-Source AI","datePublished":"2024-08-04T21:42:03+00:00","dateModified":"2024-08-21T23:22:21+00:00","mainEntityOfPage":{"@id":"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/"},"wordCount":1234,"publisher":{"@id":"https:\/\/skimai.com\/uk\/#organization"},"image":{"@id":"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#primaryimage"},"thumbnailUrl":"https:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/llama-comparison.png","articleSection":["Enterprise AI","Generative AI","LLMs \/ NLP"],"inLanguage":"pt-PT"},{"@type":"WebPage","@id":"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/","url":"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/","name":"Meta's Llama 3.1: Ultrapassar os limites da IA de c\u00f3digo aberto - Skim AI","isPartOf":{"@id":"https:\/\/skimai.com\/uk\/#website"},"primaryImageOfPage":{"@id":"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#primaryimage"},"image":{"@id":"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#primaryimage"},"thumbnailUrl":"https:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/llama-comparison.png","datePublished":"2024-08-04T21:42:03+00:00","dateModified":"2024-08-21T23:22:21+00:00","description":"Explore o Llama 3.1 da Meta, um modelo lingu\u00edstico inovador de c\u00f3digo aberto com 405 mil milh\u00f5es de par\u00e2metros. Descubra as suas funcionalidades avan\u00e7adas e as vantagens para a empresa.","breadcrumb":{"@id":"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#breadcrumb"},"inLanguage":"pt-PT","potentialAction":[{"@type":"ReadAction","target":["https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/"]}]},{"@type":"ImageObject","inLanguage":"pt-PT","@id":"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#primaryimage","url":"https:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/llama-comparison.png","contentUrl":"https:\/\/skimai.com\/wp-content\/uploads\/2024\/08\/llama-comparison.png","width":1092,"height":612,"caption":"llama comparison"},{"@type":"BreadcrumbList","@id":"https:\/\/skimai.com\/uk\/\u044f\u043a-\u043c\u0435\u0442\u0430-\u043b\u0430\u043c\u0430-3-1-\u0440\u043e\u0437\u0448\u0438\u0440\u044e\u0454-\u043c\u0435\u0436\u0456-ai-\u0437-\u0432\u0456\u0434\u043a\u0440\u0438\u0442\u0438\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/skimai.com\/"},{"@type":"ListItem","position":2,"name":"Meta&#8217;s Llama 3.1: Pushing the Boundaries of Open-Source AI"}]},{"@type":"WebSite","@id":"https:\/\/skimai.com\/uk\/#website","url":"https:\/\/skimai.com\/uk\/","name":"IA de desnata\u00e7\u00e3o","description":"A plataforma de for\u00e7a de trabalho de agentes de IA","publisher":{"@id":"https:\/\/skimai.com\/uk\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/skimai.com\/uk\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"pt-PT"},{"@type":"Organization","@id":"https:\/\/skimai.com\/uk\/#organization","name":"IA de desnata\u00e7\u00e3o","url":"https:\/\/skimai.com\/uk\/","logo":{"@type":"ImageObject","inLanguage":"pt-PT","@id":"https:\/\/skimai.com\/uk\/#\/schema\/logo\/image\/","url":"http:\/\/skimai.com\/wp-content\/uploads\/2020\/07\/SKIM-AI-Header-Logo.png","contentUrl":"http:\/\/skimai.com\/wp-content\/uploads\/2020\/07\/SKIM-AI-Header-Logo.png","width":194,"height":58,"caption":"Skim AI"},"image":{"@id":"https:\/\/skimai.com\/uk\/#\/schema\/logo\/image\/"},"sameAs":["https:\/\/www.linkedin.com\/company\/skim-ai"]},{"@type":"Person","@id":"https:\/\/skimai.com\/uk\/#\/schema\/person\/7a883b4a2d2ea22040f42a7975eb86c6","name":"Greggory Elias","url":"https:\/\/skimai.com\/pt\/author\/gregg\/"}]}},"_links":{"self":[{"href":"https:\/\/skimai.com\/pt\/wp-json\/wp\/v2\/posts\/12933","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/skimai.com\/pt\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/skimai.com\/pt\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/skimai.com\/pt\/wp-json\/wp\/v2\/users\/1003"}],"replies":[{"embeddable":true,"href":"https:\/\/skimai.com\/pt\/wp-json\/wp\/v2\/comments?post=12933"}],"version-history":[{"count":0,"href":"https:\/\/skimai.com\/pt\/wp-json\/wp\/v2\/posts\/12933\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/skimai.com\/pt\/wp-json\/wp\/v2\/media\/12938"}],"wp:attachment":[{"href":"https:\/\/skimai.com\/pt\/wp-json\/wp\/v2\/media?parent=12933"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/skimai.com\/pt\/wp-json\/wp\/v2\/categories?post=12933"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/skimai.com\/pt\/wp-json\/wp\/v2\/tags?post=12933"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}