OpenAI Unveils GPT-5.4: The Most Advanced Model for Professional Tasks Yet!

On Thursday, March 5, 2026, OpenAI launched GPT-5.4 Thinking, touted as its most sophisticated model yet for professional tasks. In ChatGPT, subscribers to Plus, Team, and Pro plans now have access to GPT-5.4 Thinking, a reasoning model that supersedes GPT-5.2 Thinking. This new model coexists with GPT-5.3 Instant, which was introduced two days earlier for conversational interactions, with an automatic mode managing the switch between the two. OpenAI also offers GPT-5.4 Pro for more intricate tasks (available only to Pro and Enterprise plans) and a standard version of GPT-5.4 in the API and Codex.

Computing, Coding, and Tool Management: What’s New in GPT-5.4

GPT-5.4, a Versatile Model Designed for Agents

GPT-5.4 combines the latest advancements from OpenAI in reasoning, coding (incorporating capabilities from GPT-5.3-Codex), and tool management. The company describes 5.4 as a model that “accomplishes complex tasks accurately and efficiently, delivering what’s requested with fewer iterations”. Significantly, it is OpenAI’s first general-purpose model that inherently includes computing abilities, meaning it can control a computer by executing clicks, keyboard commands, and interpreting screen captures, as demonstrated in the video below.

Other notable improvements include:

• A new tool-search system in the API, allowing the model to dynamically search for tool definitions instead of pre-loading them, significantly reducing token consumption.
• A context window that can extend up to a million tokens in Codex and the API.
• Improved token efficiency, with GPT-5.4 expected to solve the same issues with fewer tokens than GPT-5.2, leading to quicker responses and reduced costs.

In ChatGPT, the Thinking version now displays a preliminary work plan for complex queries, with options to adjust instructions during the response. According to OpenAI, the goal is to “more easily guide the model to the precise desired outcome without starting over or requiring additional exchanges”. The company also claims a reduction in hallucinations compared to GPT-5.2, with responses less likely to contain factual errors, as demonstrated by numerous benchmarks.

ChatGPT for Excel, Another Announcement of the Day

Alongside the launch of GPT-5.4, OpenAI unveiled the beta version of ChatGPT for Excel, a plugin that integrates ChatGPT directly into spreadsheets, similar to existing offerings like Claude. This tool allows users to build, analyze, and update models in natural language while preserving the workbook’s structure and formulas. Initially, this feature powered by GPT-5.4 is available only to Business, Enterprise, Edu, Pro, and Plus subscribers in the United States, Canada, and Australia. A version for Google Sheets is also in the works.

OpenAI Increases Launches Amid Rising Competition

This release comes just two days after the rollout of GPT-5.3 Instant, an update addressing the perceived condescension in ChatGPT’s responses noted by many users. The faster pace of releases is not coincidental.

In recent months, OpenAI has been facing a decline in its dominant position. A survey we conducted in February revealed that 54% of respondents noticed a drop in ChatGPT’s quality, and 64% of those disappointed had already switched to alternatives. Concurrently, Anthropic is leveraging this momentum: Claude recently topped the American App Store downloads and introduced Import Memory, a feature that facilitates migration from ChatGPT.

GPT-5.4 represents OpenAI’s technical response to this competitive pressure. However, the improvements claimed by the company are largely based on its internal evaluations, a metric that should be approached with caution as standardized AI model evaluations are currently facing a credibility crisis.