OpenAI launches GPT-5.4 Thinking System Card as GPT-5.4 Pro rolls out in ChatGPT
Photo by Zac Wolff (unsplash.com/@zacwolff) on Unsplash
OpenAI reports that GPT‑5.4 Pro is now live in ChatGPT, accompanied by a new “GPT‑5.4 Thinking System Card” detailing the model’s architecture and capabilities.
Key Facts
- •Key company: OpenAI
OpenAI says GPT‑5.4 Pro is now live in ChatGPT, paired with a “Thinking System Card” that outlines the model’s architecture and new capabilities. The card, posted on OpenAI’s website, details a frontier model that merges advances in reasoning, coding, and agentic workflows into a single system [OpenAI].
According to OpenAI’s rollout announcement, GPT‑5.4 is also available through the API and Codex, extending the model beyond the chat interface [OpenAI]. The company highlights a 1 million‑token context window, allowing far longer inputs than previous versions, and a 47 % reduction in token usage, which it says improves efficiency for complex tasks [report].
Benchmark data released by OpenAI shows GPT‑5.4 scoring 75 % on OSWorld‑Verified computer‑use tasks, surpassing the human baseline of 72.4 %. It also achieved 82.7 % on the BrowseComp benchmark, which measures web‑browsing and reasoning performance [report]. The company frames these results as evidence that the model is geared toward “complex knowledge work and agent workflows” rather than simple chat [report].
OpenAI’s Thinking System Card notes that the model supports mid‑generation steering, letting users interrupt and adjust responses on the fly. This “better steerability” is intended to make the model more adaptable for interactive, multi‑step processes [OpenAI]. The card also lists new agent‑style capabilities, positioning GPT‑5.4 as a tool for building autonomous workflows that can code, retrieve information, and act on instructions without constant human supervision [OpenAI].
The launch arrives amid heightened competition in the frontier‑model space. Reuters recently reported OpenAI’s earlier GPT‑5.2 release as a “code‑red” move to counter Google’s Gemini [Reuters]. TechCrunch and Ars Technica have covered OpenAI’s broader strategy of pushing agent‑mode features into productivity suites, underscoring the company’s focus on enterprise‑grade AI [TechCrunch; Ars Technica].
OpenAI’s rollout signals a shift toward larger context windows, higher efficiency, and more controllable outputs, aiming to cement its lead in the rapidly evolving AI market.
This article was created using AI technology and reviewed by the SectorHQ editorial team for accuracy and quality.