What is GPT-5.4 and what changed?
What the new model adds
OpenAI’s GPT‑5.4 is positioned as a frontier model aimed at professional and knowledge‑work use. The company released two versions — a standard GPT‑5.4 and higher‑priced GPT‑5.4 Pro — plus a specialized “Thinking” variant that emphasizes deeper reasoning and workflow orchestration. Key technical and product changes include native computer use, a much larger context window, and an improved tool‑calling mechanism intended to let models drive more complex, multi‑step tasks.
How those features translate to customers
- Native computer use: the model can interact with files, run code, and manipulate documents more directly, allowing it to perform end‑to‑end tasks instead of only producing text outputs.
- 1‑million token context: the API supports much longer sessions, enabling analysis of large documents or multi‑document synthesis without aggressive chunking.
- Better tool integration: the updated tool‑calling system makes it easier for developers to connect the model to external services and APIs.
Performance, pricing, and enterprise focus
OpenAI has claimed measurable accuracy improvements relative to its prior frontier models, saying individual claims and full responses are less likely to be false. The company also announced commercial pricing: the standard model’s stated rates were reported at roughly $2.50 per 1M input tokens and $15 per 1M output tokens, while GPT‑5.4 Pro commands substantially higher rates. The launch bundles spreadsheet plugins and other productivity features intended to appeal to finance, legal, and enterprise teams.
Why it matters
GPT‑5.4 tightens the gap between models as assistants and models as autonomous work agents. The native computer control and expanded context window make the system more useful for long, document‑heavy workflows and multi‑tool automation, accelerating adoption in professional settings — but also raising new governance, security, and cost questions for organizations that deploy it at scale.