GPT-5.4 conquers spreadsheets, transforms data analysis.

OpenAI has released GPT-5.4, its latest “frontier” model, pivoting further towards professional and enterprise applications. Designed for complex tasks like coding, data analysis, and native computer interaction, the new model aims to enhance workplace productivity. It boasts improved factual accuracy and efficiency, marking a significant step towards more autonomous AI agents capable of operating across multiple applications directly.

Key Points

• GPT-5.4 offers native computer use capabilities for multi-app task execution.
• The model significantly improves factual accuracy and reduces error generation.
• It’s optimized for professional tasks, including coding, data analysis, and presentations.
• OpenAI targets enterprise customers and developers with GPT-5.4 and higher API pricing.

OpenAI is pushing deeper into the enterprise market with the release of GPT-5.4, positioning it as their most capable model yet for professional work. This iteration builds on its predecessors by focusing on direct computer interaction and advanced reasoning, a clear signal of the company’s ambition to create more autonomous AI agents. For developers and businesses, this release could streamline workflows previously requiring complex manual oversight.

What’s New in GPT-5.4?

GPT-5.4 introduces several key advancements aimed at enhancing its utility in professional environments. Most notably, it’s OpenAI’s first model engineered with native computer-use capabilities, allowing it to execute tasks across multiple applications simultaneously. This means the model can now issue mouse and keyboard commands more effectively, significantly improving its ability to navigate a desktop environment compared to previous versions.

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.

How Does This Affect Developers and Enterprises?

For API customers, GPT-5.4 is touted as OpenAI’s most token (a unit of text processing) efficient reasoning model to date. However, this enhanced capability comes with a higher price tag. OpenAI is pricing one million input tokens at $2.50, an increase from $1.75 for GPT-5.2. This pricing adjustment underscores the model’s advanced capabilities and its target market.

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.

Why the Shift Towards Professional Work?

OpenAI’s increasing emphasis on professional applications isn’t surprising given the competitive landscape and its own financial realities. Reports last September suggested that Microsoft integrated Anthropic’s models into Copilot 365 partly because Claude was perceived as superior for tasks like generating spreadsheets and presentations. This competition has likely spurred OpenAI to refine its offerings in these critical business areas.

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.

Frequently Asked Questions

Who can access OpenAI’s GPT-5.4 model?

GPT-5.4 is primarily available to enterprise customers and developers using the OpenAI API and Codex application. It is not currently offered to Free, Go, or most Plus ChatGPT subscribers, signaling its professional focus.

How much more factual is GPT-5.4 compared to previous models?

OpenAI claims GPT-5.4 is its most factual model yet, reporting it is 18 percent less likely to generate errors in responses than GPT-5.2. Additionally, individual claims made by the model are 33 percent less likely to be false.

What new capabilities does GPT-5.4 bring for computer interaction?

GPT-5.4 is OpenAI’s first model with native computer-use capabilities, allowing it to operate across multiple applications by issuing mouse and keyboard commands. This significantly enhances its ability to navigate desktop environments and perform multi-application tasks.

What is the pricing change for GPT-5.4 API tokens?

For API customers, GPT-5.4 input tokens are priced higher than its predecessor. One million input tokens now cost $2.50, up from $1.75 for GPT-5.2, reflecting its increased capabilities and efficiency.

What This Means For You

• 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.