OpenAI Launches GPT-5.4, Embeds ChatGPT Directly Into Excel: All You Need To Know

OpenAI made a pair of closely linked and important announcements that signal a sharper focus on the workplace - a new flagship model called GPT-5.4, and a beta product that brings ChatGPT directly inside Microsoft Excel - with a particular emphasis on the financial services sector.

A new frontier model built for professional work

OpenAI released GPT-5.4 in ChatGPT (as GPT-5.4 Thinking), the API, and Codex, calling it its most capable and efficient frontier model for professional work. A more powerful variant, GPT-5.4 Pro, was also released for users who want maximum performance on complex tasks.

The company says GPT-5.4 represents a significant jump over its immediate predecessor. OpenAI reports GPT-5.4 uses 47 percent fewer tokens on some tasks than prior models, which the company says translates to faster responses and potentially lower costs for users, despite pricing being slightly higher per token than GPT-5.2.

GPT-5.4 is described as the first mainline reasoning model to incorporate the frontier coding capabilities of GPT-5.3-Codex, bringing together advances in reasoning, coding, and agentic workflows into a single model.

What's new compared to GPT-5.2

The improvements over the previous generation are measurable and wide-ranging.

On factual accuracy, GPT-5.4's individual claims are 33 percent less likely to be false and its full responses are 18 percent less likely to contain any errors relative to GPT-5.2, based on a set of de-identified prompts where users had previously flagged factual errors.

In agentic computer use - the ability for an AI to operate software on a user's behalf - GPT-5.4 reaches 75.0 percent success on OSWorld-Verified, which measures desktop navigation using screenshots plus keyboard and mouse actions, compared to 47.3 percent for GPT-5.2. Reported human performance on the same benchmark sits at 72.4 percent. This makes GPT-5.4 the first general-purpose OpenAI model to match or surpass human-level performance on that measure.

On web browsing, GPT-5.4 improves by 17 percent absolute over GPT-5.2 on BrowseComp, a benchmark measuring how well AI agents can persistently browse the web to find hard-to-locate information.

For spreadsheet work specifically - a key focus of this launch - GPT-5.4 achieves a mean score of 87.3 percent on an internal benchmark of spreadsheet modeling tasks that a junior investment banking analyst might perform, compared to 68.4 percent for GPT-5.2.

OpenAI also says GPT-5.4 outperformed office workers 83 percent of the time on GDPval, an internal benchmark measuring performance on real-world tasks across 44 occupations.

ChatGPT comes to Excel

The headline product announcement paired with GPT-5.4 is ChatGPT for Excel - a native add-in that brings OpenAI's AI directly into workbooks, rather than requiring users to switch between applications.

The ChatGPT for Excel add-in, now in beta, brings ChatGPT directly into workbooks to help build and update models, run scenarios, and generate outputs based on cells and formulas. It is powered by GPT-5.4.

The tool allows teams to describe what they need in plain language, with ChatGPT then creating or updating live Excel models directly in the workbook - covering use cases from data analysis and reporting to inventory management and budgeting, while preserving structure, formulas, and assumptions in a formatted, Excel-native workbook.

OpenAI also touts the tool's ability to reason across entire workbooks. ChatGPT can understand how sheets and formulas connect across a model, explain why outputs changed, trace and fix errors, and show how assumptions flow through a model - capabilities especially useful when users inherit existing templates or want to test a workbook before making decisions.

A push Into financial services

Both announcements are tied to a broader OpenAI push into financial services. OpenAI is also adding financial data integrations directly in ChatGPT for FactSet, Dow Jones Factiva, LSEG, Daloopa, S&P Global, and more, making it easier to work with trusted financial data inside ChatGPT.

On the internal investment banking benchmark, performance improved from 43.7 percent with GPT-5 to 87.3 percent with GPT-5.4 Thinking - evaluating real-world workflows such as building a three-statement financial model with proper formatting and citations.

OpenAI is also introducing reusable "Skills" for recurring finance work such as earnings previews, comparables analysis, DCF analysis, and investment memo drafting.

Availability

GPT-5.4 Thinking is available starting immediately to ChatGPT Plus, Team, and Pro users, replacing GPT-5.2 Thinking. GPT-5.2 Thinking will remain available for three months for paid users under the Legacy Models section before being retired on June 5. GPT-5.4 Pro is available to Pro and Enterprise plans.

Published on: Friday, March 06, 2026, 10:59 AM IST