OpenAI Unveils GPT-5.4 with Native Computer Use Mode and Financial Plugins for Excel & Google Sheets, Boosting Efficiency and Automation
OpenAI has launched GPT-5.4, a significant upgrade following GPT-5.3 Instant. This new model comes in two versions: GPT-5.4 Thinking and GPT-5.4 Pro, with Thinking available to all paid ChatGPT subscribers and Pro reserved for ChatGPT Pro and Enterprise users. A key highlight is its improved efficiency, using 47% fewer tokens on some tasks. More notably, GPT-5.4 introduces a 'native' Computer Use mode via API and Codex, allowing it to navigate and operate across a user's computer like a human. Additionally, new integrations enable GPT-5.4 to plug directly into Microsoft Excel and Google Sheets for granular analysis and automated task completion, potentially accelerating enterprise work but also raising concerns about white-collar job displacement. GPT-5.4 supports up to 1 million tokens of context in the API and Codex, facilitating long-horizon task planning and execution, though costs double for inputs exceeding 272,000 tokens.
The pace of AI updates continues unabated, with OpenAI launching GPT-5.4 just two days after its GPT-5.3 Instant model. This latest, more substantial upgrade is offered in two distinct varieties: GPT-5.4 Thinking and GPT-5.4 Pro. GPT-5.4 Thinking will be accessible to all paid subscribers of ChatGPT, including the $20-per-month Plus plan and higher tiers. GPT-5.4 Pro, designed for the most intricate tasks, will be exclusively available to ChatGPT Pro users, who pay $200 monthly, and those on Enterprise plans. Both versions will also be integrated into OpenAI's paid application programming interface (API) and Codex software development application. ChatGPT Free users will also have limited access to GPT-5.4, but only when their queries are automatically routed to the model, as confirmed by an OpenAI spokesperson.
Key features of this release emphasize efficiency and advanced functionality. OpenAI reports that GPT-5.4 demonstrates significantly improved efficiency, utilizing 47% fewer tokens on certain tasks compared to its predecessors. Perhaps even more impressively, the update introduces a new "native" Computer Use mode. Available through the API and Codex, this mode empowers GPT-5.4 to navigate a user's computer and operate across various applications in a manner akin to a human user. This capability marks a significant step towards more autonomous workflows.
Furthermore, OpenAI is rolling out a new suite of ChatGPT integrations. These integrations allow GPT-5.4 to be directly embedded into users' Microsoft Excel and Google Sheets spreadsheets and individual cells. This enables granular data analysis and automated task completion, which is expected to accelerate work processes across enterprises. However, this advancement, following similar offerings like Anthropic's Claude and its new Cowork application, may intensify concerns regarding potential white-collar job displacement.
In terms of capacity, OpenAI states that GPT-5.4 supports up to 1 million tokens of context within its API and Codex. This extensive context window allows AI agents to plan, execute, and verify complex tasks over extended horizons. It's important to note, however, that the cost per 1 million tokens doubles once the input exceeds 272,000 tokens. The introduction of native computer use is highlighted by OpenAI as its most consequential capability, signaling a move towards more autonomous and integrated AI systems.