OpenAI has officially launched GPT-5.4, a groundbreaking frontier model that brings native computer use, advanced reasoning, and elite coding capabilities into a single package. The highly anticipated OpenAI GPT-5.4 release marks a major shift in artificial intelligence, as the model can now independently operate software, analyze complex spreadsheets, and power long-horizon professional workflows. For everyday users and developers alike, this update makes every other AI model look slow by comparison.
The Rise of Native Computer Use
The most significant change with the OpenAI GPT-5.4 upgrade is its ability to take direct control of a computer. Unlike previous versions, GPT-5.4 can navigate software by clicking, typing, and responding to screenshots without relying on specialized secondary models.
On the OSWorld-Verified benchmark, which tests how well an AI navigates a real desktop environment, GPT-5.4 scored 75.0 percent. This massive leap crushes the 47.3 percent score of its predecessor, GPT-5.2, and even edges past the measured human baseline of 72.4 percent. Developers can now build autonomous agents that actively operate websites and applications rather than just generating text about them.
A Massive Upgrade for Professional Workflows
OpenAI specifically engineered GPT-5.4 to excel at real-world professional tasks. The model is highly proficient at building financial models, drafting legal documents, and editing presentations.
When tested on spreadsheet modeling tasks designed for junior investment banking analysts, GPT-5.4 scored an impressive 87.5 percent. This is a significant jump from the 68.4 percent achieved by GPT-5.2. Furthermore, human evaluators preferred the visual variety and image generation of GPT-5.4’s presentations over the previous model 68 percent of the time.
Accuracy has also seen a major improvement. Individual factual claims made by GPT-5.4 are 33 percent less likely to be false compared to GPT-5.2. Additionally, full responses are 18 percent less likely to contain any errors, providing a much-needed reliability boost for professionals handling critical data.
Faster and Smarter Coding Capabilities
With OpenAI GPT-5.4, the model now serves as the primary coding engine, eliminating the need to choose between ChatGPT and Codex for most programming tasks. It matches or outperforms the specialized GPT-5.3-Codex on the SWE-Bench Pro benchmark while operating significantly faster.
A new fast mode in Codex delivers up to a 1.5x speed improvement across all supported models. GPT-5.4 is particularly skilled at complex front-end coding, producing functionally correct and aesthetically polished results. OpenAI also introduced an experimental feature called “Playwright (Interactive),” which allows Codex to visually debug web and Electron applications in real time, actively testing the apps as it builds them.
ChatGPT 5.4 Thinking and Everyday Usage
For regular users, the most noticeable update is the introduction of ChatGPT 5.4 Thinking. When faced with a complex, multi-step prompt, the chatbot now displays an upfront plan before it starts generating a response. This “thinking out loud” approach allows users to intervene and adjust the model’s direction mid-response if the plan misses the mark.
For instance, if asked to plan a family vacation, the AI might outline its intent to check flights, estimate costs, and build an itinerary. Users can instantly correct the plan to prioritize a road trip instead. The model also maintains much stronger contextual awareness over extended conversations.
While ChatGPT 5.4 Thinking handles deep, analytical tasks, ChatGPT 5.3 Instant remains the default model for everyday, fast-paced conversations. This layered approach ensures users get the right balance of speed and computational power when they need it.
Availability and Improved Efficiency for Developers
The new features are rolling out across multiple tiers. ChatGPT Plus, Team, and Pro users can now access GPT-5.4 Thinking. Enterprise and Edu plan administrators can enable early access through their settings. Meanwhile, a more powerful GPT-5.4 Pro version is exclusively available to Pro and Enterprise subscribers for highly demanding analytical work.
Developers can immediately access the new models via the API. One of the biggest backend improvements is a new tool-search configuration. Instead of loading every available tool’s full definition into the context window, the model receives a lightweight list and looks up specific tools only when needed. In testing on the MCP Atlas benchmark, this efficiency update reduced total token usage by 47 percent while maintaining accuracy, translating to faster response times and lower costs for developers.
