Pular para o conteúdo
Menu
Mix Vale

OpenAI leaks detail GPT-5.4, promising extreme reasoning and context expansion

Carolina Costa
Por 6 min de leitura
Compartilhe
chatgpt health
Foto: chatgpt health - Amir Sajjad/Shutterstock.com

The global technology scene has been roiled by a cryptic announcement and the discovery of quickly removed code snippets pointing to the imminent release of GPT-5.4. The next generation of OpenAI’s language model is already the focus of intense speculation, promising to redefine the capabilities of generative artificial intelligence.

Publicidade

This update, which at first glance might seem incremental, proves to be a potential game-changer in the fierce race for AI supremacy. With expected features such as an “Extremo Reasoning” mode and a significantly expanded context window, OpenAI seeks to consolidate its leadership over strong competitors.

Accelerated development and a more frequent release cycle, however, raise important discussions about rising infrastructure costs and ethical implications. The economic viability and social responsibility of progress in AI are increasingly on the agenda as models become more autonomous and versatile.

Publicidade

Unmistakable signs: the involuntary revelation of GPT-5.4

Public confirmation of GPT-5.4 came not through a formal press release, but through a series of unexpected events. On March 3, an hour after the release of GPT-5.3 Instant, OpenAI published a message of just five words on its social media platform, generating millions of views and intense discussion. The phrase “5.4 sooner than you think”, without any other explanation, was the catalyst for the developer and investor community.

This cryptic message did not emerge in isolation. Ela was preceded by a week of leaks in OpenAI’s Codex repository, where snippets of code revealed internal aspects of the model under development before being promptly deleted. Tais incidents, common in the software development process, provided concrete evidence of the existence and planned features of GPT-5.4, fueling expectations for what could be a significant leap forward in AI technology.

Context Window Expansion: Processing Vast Volumes

One of the most impactful features of GPT-5.4, as indicated by the leaks, is the substantial enlargement of its context window. Análises of commits reveal the possibility of a capacity of up to two million tokens, which represents a significant increase compared to previous models in the GPT-5 and Codex GPT-5.3 series. Para illustrate, this capacity would be equivalent to processing approximately five thousand printed pages, allowing the AI ​​to handle entire codebases, extensive legal cases, or voluminous scientific articles in a single work session. Essa capability redefines the limits of what is viable in terms of analyzing complex data, paving the way for applications in areas that require the handling of large volumes of information without the need for fragmentation, a crucial advance for the efficiency and depth of interactions with artificial intelligence.

Extreme reasoning: deepening analysis through intelligence

In addition to expanding the context window, GPT-5.4 promises an “Extreme” reasoning mode, according to reports from sources close to OpenAI’s plans. Essa functionality will allow the model to allocate a significantly greater amount of computing power to solving complex questions, resulting in more in-depth and accurate cognitive analyses.

This mode is designed mainly for researchers and for solving highly complex scientific, mathematical and technical problems, differentiating itself from the agility expected by common users. The expectation is that this capability will transfer part of the computational effort from the training phase to the inference phase, allowing the model to invest more resources in formulating the final answer.

Publicidade

Precision image analysis: a new level for applications

A third technical innovation, evidenced by leaked pull requests, focuses on improving image processing. The code in question adds a feature that allows the direct and uncompressed passage of original image data in PNG, JPEG and WebP formats to the responses API. Esta functionality, controlled by a new parameter, “detail: original”, has a minimum requirement of version 5.4, indicating that it is exclusive to this model.

Currently, existing GPT models compress images before processing, which can compromise analysis quality in tasks that require pixel-level precision. Isso affects crucial areas such as interpreting medical images, analyzing satellite data, optical character recognition in legal or financial documents, and reviewing architectural plans and technical schematics. The ability to process images at full resolution will propel GPT-5.4 into a new range of professional applications where loss of detail has been a limiting factor.

For sectors such as manufacturing, which depend on quality assurance through artificial intelligence, or the legal and financial segments, which automate document processing, and medicine, which uses image diagnostics, this innovation represents a considerable advance. The explicit binding of this feature to GPT-5.4 is no accident, as processing uncompressed images requires substantially more computational power and memory bandwidth, increasing technical demands on the model and the entire underlying infrastructure.

Strategic acceleration: the intense pace of OpenAI releases

The discussion around GPT-5.4 also highlights the increasing speed at which OpenAI has been releasing new variants of its models. Desde the launch of GPT-5 in August, the company has maintained an unprecedented pace of updates, surpassing the release frequency of the GPT-4 era.

The recent timeline demonstrates this acceleration: GPT-5.1 was released in November, GPT-5.2 in December, and several variants of GPT-5.3, including Codex and Instant, were made available in February and March. Caso GPT-5.4 is released as scheduled in March or April, the interval between versions would be just one month, consolidating a rapid release strategy.

This strategy, according to internal sources, is a deliberate decision to manage user expectations and maintain market engagement. Lançamentos more frequent incremental improvements aim to maintain the sector’s attention without the risk of a single major event that may disappoint. The high expectations surrounding GPT-5 demonstrated the difficulty of exceeding projections, and the current approach seeks continued and more stable growth of the user base.

Implications for the artificial intelligence ecosystem

The rapid advancement of OpenAI with models like GPT-5.4 brings with it a series of implications for the entire artificial intelligence ecosystem. Competition between technology giants like Google and Anthropic intensifies, driving innovation and the development of even more sophisticated capabilities. However, the fast pace also raises concerns about operational costs and fatigue for developers who need to constantly adapt their applications to new model versions.

Publicidade

Carolina Costa é colaboradora da redação do Mix Vale, portal de notícias com cobertura de Brasil e mundo. Produz conteúdo nas editorias de economia e benefícios, com foco em informação de serviço para o leitor brasileiro. Escreve para o portal desde 2019. Formação: Administração

Notícias relacionadas

Mais lidas

  1. Ciência

    Meteoro no Nordeste: bólido cruza o céu de cinco estados às 21h34 e mobiliza buscas por fragmentos na Paraíba

  2. Caixa Econômica Federal

    Pé-de-Meia paga parcela de R$ 1.000,00 a aprovados no ensino médio de 2026 entre fevereiro e março de 2027

  3. Política

    Com obstrução parcial no intestino, Jair Bolsonaro é internado no hospital DF Star em Brasília

  4. INSS

    Recuperação da senha Gov.br pelo celular exige reconhecimento facial ou código por SMS para o Meu INSS

  5. EUA

    Terremoto em Los Angeles atinge magnitude 4,2 e causa tremores leves em Commerce na madrugada

  6. Matéria Publicitária

    MineMV é o novo jogo grátis de construir com blocos; veja como jogar e liberado para todas as idades

  7. Política

    Regulamentação da escala 5×2 pelo governo encarece cálculo da hora extra a partir de 16 de dezembro

  8. Santa Catarina

    Acidente de trânsito na SC-108 envolve carro e moto e deixa motociclista morto em São Ludgero

  9. Mundo

    Jornalista italiano Corrado Augias morre aos 91 anos; presidente da Itália lamenta a perda

  10. Loterias

    Rendimento de R$ 40 milhões da Mega-Sena 3088 chega a R$ 337 mil por mês no Tesouro Direto