OpenAI has released GPT-5.4, its most capable and efficient model for professional and agentic work to date. Available across ChatGPT, the developer API, and Codex, the release marks a decisive step in OpenAI’s strategy to evolve from a language model provider into an autonomous AI platform.
Native Computer Use
The headline capability of GPT-5.4 is native computer use — the model can autonomously operate applications, navigate web browsers, fill in forms, and execute complex multi-step workflows across software tools by reacting to screenshots and issuing keyboard and mouse commands.
This is no longer a bolt-on feature. Computer use is built into the core model, tightly integrated with Codex for developer workflows and available directly in ChatGPT for professional users.
Context Window and Codex Integration
GPT-5.4 ships with:
- 272,000 token standard context window
- 1 million token experimental context window — available to developers through Codex
- Full integration of the coding strengths previously found in GPT-5.3-Codex
For engineering teams, the 1M context window is transformative: entire codebases, documentation sets, and long conversation histories can now be loaded in a single prompt.
Mini and Nano Variants
Shortly after the flagship release, OpenAI introduced GPT-5.4 mini and GPT-5.4 nano — smaller, faster variants optimised for high-volume workloads. GPT-5.4 mini runs more than twice as fast as its predecessor while improving across coding, reasoning, multimodal understanding, and tool use.
Why It Matters
GPT-5.4 cements the trajectory OpenAI has been building toward: AI that does not just answer questions but executes work end-to-end. Combined with Codex’s development environment integration, GPT-5.4 is positioned as a first-class autonomous collaborator for knowledge workers and software engineers alike.
The race to build the AI that works for you — not just with you — just got a meaningful benchmark to beat.
Source: openai.com, gizmodo.com, aiinsider.in