What’s new in OpenAI’s GPT‑5.4 for professional work?
How the update changes AI for knowledge workers
OpenAI’s GPT‑5.4 is positioned as a leap toward models built for professional, knowledge‑work tasks. The company released two variants—GPT‑5.4 Thinking, which emphasizes deeper reasoning, and GPT‑5.4 Pro—alongside API access and integrations that target office workflows.
Major technical and product changes
- Native computer use: the model can interact with software environments more directly, allowing it to run tasks that previously required manual tool orchestration.
- Huge context windows: OpenAI announced support for context windows up to 1 million tokens in API use, enabling much larger documents and extended workflows to be processed in a single session.
- Improved tool calling: the API includes more robust tool‑calling capabilities so GPT‑5.4 can coordinate external plugins and workflow steps with fewer errors.
Work‑focused features
- Spreadsheets and finance plugins: new integrations let the model operate inside Microsoft Excel and Google Sheets, plus a set of financial‑services tools aimed at accounting and analysis tasks.
- Professional benchmarks: OpenAI says GPT‑5.4 sets new records on a suite of industry benchmarks and real‑world occupational tests, with independent reports noting strong gains on multi‑disciplinary tasks.
Safety, accuracy, and pricing
OpenAI reported that the model’s individual claims are materially less likely to be false compared with prior releases, and metrics shared by the company indicate reductions in whole‑response error rates versus GPT‑5.2. Pricing details were published for API use—different per‑token rates for the standard and Pro tiers—which reflects the company’s positioning of the offering for enterprise and heavier professional workloads.
Why it matters
The release tightens the gap between narrow productivity tools and general‑purpose models. By focusing on native computer control, larger context, and built‑in spreadsheet workflows, OpenAI is aiming to make LLMs more directly useful to analysts, finance teams, and software engineers—and to accelerate enterprise adoption.