Connect with us

Hi, what are you looking for?

AI Generative

OpenAI Launches GPT-5.4: 33% Fewer Errors and Enhanced Real-World Task Performance

OpenAI launches GPT-5.4 with 33% fewer errors and a state-of-the-art 83% success rate in real-world tasks, enhancing productivity for professionals.

OpenAI announced the release of GPT-5.4 this week, touting significant improvements for users of its ChatGPT platform. The launch follows closely on the heels of GPT-5.3 Instant, a model designed with everyday users in mind. OpenAI describes GPT-5.4 as the “most capable and efficient frontier model for professional work,” indicating a clear focus on enhancing productivity for professionals and developers. The company also introduced two new features: GPT-5.4 Thinking and GPT-5.4 Pro, aimed at those seeking optimal performance.

Among the notable enhancements is a marked improvement in efficiency. OpenAI claims that GPT-5.4 is the “most token-efficient reasoning model yet,” utilizing significantly fewer tokens to address problems compared to its predecessor, GPT-5.2. This increase in efficiency translates into faster response times and reduced token consumption, which could lead to cost savings for users relying on token-based pricing models.

Moreover, the new model demonstrates enhanced capabilities in executing real-world tasks pertinent to knowledge work. OpenAI reported that, in evaluations conducted on GDPval, a benchmark that measures agents’ abilities to perform well-defined knowledge tasks across 44 occupations, GPT-5.4 achieved a new state of the art. It outperformed its predecessor, matching or exceeding industry professionals in 83.0% of comparisons, a substantial increase from the 70.9% success rate recorded by GPT-5.2. This improvement suggests that GPT-5.4 is now well-positioned to perform tasks traditionally held by human professionals.

Accuracy is another area where OpenAI claims to have made strides. The company asserts that GPT-5.4 is 33% less likely to generate false claims and 18% less likely to produce erroneous statements when compared to GPT-5.2. Such enhancements could help bolster user trust in the information generated by the model and expand its applicability across various sectors.

In addition to these improvements, GPT-5.4 introduces a more flexible approach to user interaction. The model allows users to adjust its responses mid-explanation, providing a more collaborative experience. OpenAI stated that in ChatGPT, the new GPT-5.4 Thinking feature can present a preliminary roadmap of its reasoning process, enabling users to refine the output as it is being generated. This real-time adjustment capability could prove particularly beneficial in professional settings where precision and relevance are critical.

OpenAI is also catering to enterprise clients with the launch of a new plugin, the ChatGPT for Excel add-in. This tool is designed to enhance the model’s ability to generate and analyze spreadsheets, further integrating AI capabilities into daily business workflows. Such innovations could streamline data management processes and improve decision-making efficiency for organizations.

The rollout of GPT-5.4 begins today, and it will be accessible through ChatGPT, Codex, and the OpenAI API. Currently, users will need a paid ChatGPT subscription to access the new model, although this may change in the future as OpenAI continues to evolve its offerings.

As AI technology continues to advance rapidly, the introduction of GPT-5.4 represents a significant step forward in enhancing the capabilities of generative models for professional applications. OpenAI’s focus on efficiency, accuracy, and real-world task execution highlights a growing trend toward integrating AI into the workforce. The implications of these advancements could reshape not only individual job roles but also entire industries, as AI becomes increasingly adept at tackling complex tasks traditionally handled by humans.

For further details, users can refer to OpenAI’s official blog post regarding the launch.

Disclosure: In April 2025, Ziff Davis, the parent company of Mashable, filed a lawsuit against OpenAI, alleging copyright infringement in the training and operation of its AI systems.

See also
Staff
Written By

The AiPressa Staff team brings you comprehensive coverage of the artificial intelligence industry, including breaking news, research developments, business trends, and policy updates. Our mission is to keep you informed about the rapidly evolving world of AI technology.

You May Also Like

Top Stories

Intern Chris Maddison's pivotal contributions to Google DeepMind's AlphaGo led to its historic victory over Go champion Lee Sedol, revolutionizing AI's capabilities.

AI Research

Researchers unveil Humanity's Last Exam, revealing top AI models like OpenAI's GPT-4 and Claude scored just 2.7% to 3.5%, highlighting significant limitations.

AI Marketing

Google's search revenue surged 16% to $102.3 billion amid AI advancements, as user queries grow, despite a 54.6% drop in organic click-through rates.

Top Stories

OpenAI unveils GPT-5.2 with a groundbreaking 400,000-token context window and enhanced reasoning, despite a 39% hallucination rate, reshaping AI capabilities.

Top Stories

Anthropic, after losing a $200M DOD contract, sees a surge in success with over 1M daily downloads of its Claude app, emphasizing ethical AI...

AI Regulation

OpenAI's new contract with the Pentagon raises alarms over potential surveillance use of its technology, igniting protests and calls for ethical accountability.

AI Technology

TU Braunschweig unveils a groundbreaking GaN microLED system aimed at revolutionizing AI chip efficiency, potentially reducing energy consumption by up to 40% by 2035.

AI Government

Hackers exploited ChatGPT and Claude to exfiltrate 150GB of sensitive data from the Mexican government, compromising 195 million taxpayer records.

© 2025 AIPressa · Part of Buzzora Media · All rights reserved. This website provides general news and educational content for informational purposes only. While we strive for accuracy, we do not guarantee the completeness or reliability of the information presented. The content should not be considered professional advice of any kind. Readers are encouraged to verify facts and consult appropriate experts when needed. We are not responsible for any loss or inconvenience resulting from the use of information on this site. Some images used on this website are generated with artificial intelligence and are illustrative in nature. They may not accurately represent the products, people, or events described in the articles.