In ChatGPT's Thinking mode, where GPT-5.4 will become the default, users will see the system outline its planned approach to a request. This provides a critical feedback loop, allowing users to adjust the model's course during response generation. OpenAI also claims enhanced web research capabilities, particularly for "highly specific" queries, asserting that the model "can more persistently search across multiple rounds to identify the most relevant sources, particularly for 'needle-in-a-haystack' questions, and synthesize them into a clear, well-reasoned answer."
Factual accuracy has also seen a significant bump. OpenAI states that GPT-5.4 is its "most factual model yet," being 18 percent less likely to generate errors compared to GPT-5.2. Furthermore, individual claims are 33 percent less likely to be false. These improvements are designed to yield "higher-quality answers that arrive faster and stay relevant to the task at hand," according to the company.
GPT-5.4 will be available to enterprise customers and developers leveraging the company's Codex app, as well as via the API. This means it won't be immediately accessible to Free or Go users, nor even most Plus subscribers. This strategic rollout emphasizes OpenAI's focus on professional-grade applications and solutions for businesses willing to invest in cutting-edge AI. The company is also debuting OpenAI for Financial Services, which includes a version of ChatGPT that can operate directly within spreadsheets.
Reporting from The Information suggests OpenAI is generating approximately $25 billion in annualized revenue. However, the company faces substantial financial commitments, with over $1.4 trillion in data center obligations on its books. This makes investor funding crucial for its operations. By focusing on productivity solutions, OpenAI aims to carve out a sustainable business model in a market segment willing to pay for advanced, reliable AI tools.
For Developers
The increased token efficiency, despite the higher cost, suggests that complex, multi-step agentic workflows could become more viable and performant, potentially reducing overall compute for intricate tasks. Explore integrating native computer-use capabilities into your applications for truly autonomous agents.
For Founders and Product Managers
The enhanced factual accuracy and reduced error rate of GPT-5.4 could significantly lower the risk associated with deploying AI for critical business functions like data analysis or automated report generation. Consider building products that leverage its improved desktop navigation for tasks currently requiring human oversight.
For Enterprise Leaders
With GPT-5.4's focus on professional tasks and native computer interaction, evaluate opportunities to automate cross-application workflows within your organization. The "Thinking mode" transparency could be crucial for compliance and oversight in regulated industries, allowing teams to monitor and adjust AI behavior.
GPT-5.4 is OpenAI's latest AI model designed for enterprise use, focusing on tasks like coding, data analysis, and computer interaction. Key features include native computer use capabilities for executing tasks across multiple applications, improved factual accuracy with 18% fewer errors than GPT-5.2, and enhanced web research for specific queries.
GPT-5.4 significantly improves factual accuracy, with an 18% reduction in error generation compared to GPT-5.2. Individual claims made by the model are also 33% less likely to be false, leading to higher-quality and more reliable answers.
GPT-5.4 is designed to streamline workflows for developers and enterprises by enabling more autonomous AI agents. It can execute tasks across multiple applications, improve reasoning, and enhance web research capabilities, all while being more token-efficient for API usage.
GPT-5.4 has enhanced web research capabilities, especially for highly specific queries, and can persistently search across multiple rounds to identify relevant sources. It synthesizes information into clear, well-reasoned answers, making it effective for 'needle-in-a-haystack' questions.
More insights on trending topics and technology
