OpenAI Unveils GPT-5.4: A Leap Forward in AI Reasoning and Automation
On March 5, 2026, OpenAI introduced GPT-5.4, its most advanced AI model to date, integrating enhanced reasoning, coding proficiency, and autonomous workflow capabilities into a unified system. This release signifies a major advancement in artificial intelligence, offering users a more efficient and versatile tool for a wide range of applications.
Unified System with Enhanced Capabilities
GPT-5.4 consolidates features that were previously distributed across multiple models. It combines the superior coding abilities of GPT-5.3-Codex with improved general reasoning and native computer-use functionalities. This integration enables the model to handle complex tasks more effectively, reducing the need for extensive user interaction. From managing spreadsheets and creating presentations to executing intricate multi-step processes, GPT-5.4 streamlines professional workflows.
Improved User Interaction in ChatGPT
Incorporated into ChatGPT as GPT-5.4 Thinking, the model introduces an upfront reasoning plan. This feature allows users to interrupt and redirect the model mid-response without restarting the entire process, facilitating more precise and contextually accurate outputs. This real-time steerability marks a significant improvement over previous models, where course corrections required beginning anew.
Benchmark Performance
GPT-5.4 sets new industry standards across several critical benchmarks:
– GDPval: The model matches or exceeds industry professionals in 83% of comparisons across 44 occupations spanning the top nine U.S. GDP industries, an increase from 70.9% with GPT-5.2.
– SWE-Bench Pro (Public): Achieves a score of 57.7%, surpassing previous iterations.
– OSWorld-Verified: Attains a 75.0% success rate, exceeding human performance benchmarked at 72.4% and significantly outperforming GPT-5.2’s 47.3%.
– Toolathlon: Scores 54.6%, reflecting enhanced tool usage capabilities.
– BrowseComp: Achieves an 82.7% success rate, indicating improved browsing comprehension.
In the legal domain, GPT-5.4 scored 91% on the BigLaw Bench evaluation for legal document work, as reported by Harvey’s Head of Applied Research, Niko Grupen.
Native Computer-Use Capabilities
GPT-5.4 is OpenAI’s first general-purpose model with native computer-use capabilities. It enables agents to interact directly with software through screenshots, mouse commands, and keyboard inputs. This functionality allows the model to perform tasks that require direct manipulation of computer interfaces, enhancing its applicability in various professional settings.
On the OSWorld-Verified benchmark, GPT-5.4 achieves a 75.0% success rate, surpassing human performance and significantly outperforming GPT-5.2. In the WebArena-Verified benchmark, it attains a 67.3% browser success rate and scores 92.8% on Online-Mind2Web using screenshot-based observations alone.
Extended Context Window
The model supports a context window of 1 million tokens in the API, enabling it to handle long-horizon tasks across large-scale workflows. This extended context window matches offerings from competitors like Google and Anthropic, allowing GPT-5.4 to maintain coherence and relevance over extended interactions.
Enhanced Accuracy and Efficiency
OpenAI emphasizes that GPT-5.4 is its most factual model yet. Individual claims are 33% less likely to be false, and full responses are 18% less likely to contain errors compared to GPT-5.2. Additionally, the model delivers significant token-efficiency gains, using substantially fewer tokens to solve the same reasoning problems. This efficiency translates directly into reduced API costs and faster response times for enterprise developers.
In production environments, Mainstay CEO Dod Fraser reported that GPT-5.4 achieved a 95% first-attempt success rate across approximately 30,000 property portals. It completed sessions three times faster while using 70% fewer tokens compared to prior models.
Availability
GPT-5.4 Thinking is now available for ChatGPT Plus, Team, and Pro subscribers, replacing GPT-5.2 Thinking over the next three months. Developers can access GPT-5.4 and GPT-5.4 Pro through the OpenAI API, with priority processing enabled for faster token velocity in production environments.
