TITLE: GPT-5.4 Can Now Control Your Computer Autonomously DATE: 2026-03-13 COMPANY: OpenAI TOPIC: Model Releases SUMMARY: OpenAI released GPT-5.4 on 5 March 2026, the first general-use AI model with native computer-use capabilities. The model surpasses the human benchmark for real-world computer tasks and embeds directly into Excel and Google Sheets, bringing autonomous workflow execution to everyday business tools. WHAT CHANGED: OpenAI released GPT-5.4 on 5 March 2026, describing it as its "most capable and efficient frontier model for professional work." The release combines advanced reasoning, coding, and autonomous computer operation into a single model, available in three versions: GPT-5.4 Standard, GPT-5.4 Pro, and GPT-5.4 Thinking. The headline capability is computer use. GPT-5.4 is the first general-use OpenAI model with native computer-use built in, meaning it can navigate operating systems, browsers, and software applications without requiring custom integrations from developers. On OSWorld-Verified, a standardised benchmark for real-world computer tasks, GPT-5.4 achieves a 75.0% success rate. The human benchmark sits at 72.4%. Its predecessor, GPT-5.2, scored 47.3% on the same test. On WebArena-Verified, it achieves a 67.3% browser task success rate. Alongside the model, OpenAI launched ChatGPT for Excel and Google Sheets in beta. The integration embeds ChatGPT directly into spreadsheet applications, allowing teams to build, analyse, and update complex financial models without leaving familiar tools. New data integrations with FactSet, MSCI, Third Bridge, and Moody's allow teams to pull live market and company data into their workflows from within the same interface. The model supports a 1 million token context window via the API, matching context capacity offered by Google and Anthropic. OpenAI also reports that GPT-5.4 is its most factual model to date: individual claims are 33% less likely to be false, and full responses are 18% less likely to contain errors, compared to GPT-5.2. WHY IT MATTERS: Computer-use AI crossing the human benchmark is a threshold moment. Autonomous task execution across real applications is no longer theoretical. Small teams can now automate multi-step, multi-application workflows without engineering resources or custom integrations. The Excel and Google Sheets integration brings AI-assisted financial modelling directly into existing tools, lowering adoption friction for finance and operations teams. Live data integrations with financial information providers mean AI can pull, analyse, and report on external data inside a single workflow. Lower hallucination rates make GPT-5.4 more viable for compliance-sensitive and client-facing use cases where factual accuracy is non-negotiable. The 1 million token context window enables long-horizon task execution across large datasets and complex, multi-step agent workflows. DAVID & GOLIATH ANALYSIS: The computer-use benchmark result matters beyond the number. When an AI model can outperform a human on real-world computer tasks, including navigating real software on a real operating system, the category of "things AI can automate" expands significantly. Operators who have been waiting for AI to handle genuinely complex, multi-step workflows should note that the technical threshold has now been crossed. The Excel and Google Sheets integration deserves particular attention for smaller operators. Most finance, operations, and admin work happens inside spreadsheets. An AI that can sit inside those tools, pull live data from professional information services, and build or update models without requiring a developer closes a gap that previously required either dedicated technical staff or expensive enterprise software. The practical recommendation is to map your highest-frequency, highest-friction workflows and ask whether they involve navigating multiple applications or maintaining complex spreadsheet models. Those are the workflows GPT-5.4 is now capable of handling. Start with one. Measure the time saving. Scale from there. RELEVANT SYSTEMS: AI Growth Engine, Employee Amplification Systems SOURCE URL: https://davidandgoliath.ai/daily-ai-briefing/openai-gpt-54-computer-use-beats-human-benchmarks FEED URL: https://davidandgoliath.ai/daily-ai-briefing/feed --- Published by David & Goliath | https://davidandgoliath.ai Daily AI Briefing: one AI development per day, decoded for business operators. This is a structured companion file optimised for LLM retrieval and citation.