OpenAI launches GPT-5.4 Thinking and Pro combining coding, reasoning, and computer use in one model

2026-03-05 · Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering, Cybersecurity & Data Privacy · Depth: Intermediate, medium

Summary

OpenAI has launched GPT-5.4 Thinking and GPT-5.4 Pro, integrating coding, reasoning, agentic workflows, and native computer operation into a single model for the first time. The new model achieved an 83.0 percent score on the GDPval benchmark for professional knowledge work, a notable increase from GPT-5.2's 70.9 percent, and surpassed human performance on the OSWorld Verified benchmark for desktop environment navigation with a 75.0 percent success rate. A key technical enhancement, "Tool Search" in the API, reduces token consumption by 47 percent by retrieving tool definitions only when needed. While coding gains are modest on SWE-Bench Pro, a new "/fast" mode in Codex boosts token speed by up to 1.5x. GPT-5.4 also features improved visual perception, reduced hallucinations, and a "High Capability" cybersecurity rating.

Key takeaway

For AI architects evaluating new frontier models, GPT-5.4's integration of coding, reasoning, and native computer use, alongside its "High Capability" cybersecurity rating, presents a compelling option for enterprise applications. You should consider its improved performance on professional benchmarks and token efficiency gains, despite increased per-token pricing, for automating complex workflows and enhancing agentic systems.

Key insights

GPT-5.4 unifies coding, reasoning, and computer operation, significantly boosting performance in professional knowledge work and desktop navigation.

Principles

• Dynamic tool loading optimizes token efficiency.
• Integrated capabilities enhance general model utility.

Method

GPT-5.4 employs "Tool Search" to dynamically load tool definitions, reducing token consumption. It also features a "/fast" mode for accelerated coding and a two-stage monitoring system for cybersecurity.

In practice

• Utilize "Tool Search" for large tool ecosystems.
• Employ "/fast" mode for faster coding tasks.
• Leverage GPT-5.4 for complex office automation.

Topics

• GPT-5.4
• Large Language Models
• Agentic AI
• Token Efficiency
• Cybersecurity

Best for: Machine Learning Engineer, CTO, AI Architect, AI Engineer, Data Scientist, AI Product Manager

Related on AIssential

• Introducing GPT-5.4 — GPT-5.4 enhances professional AI capabilities through integrated reasoning, coding, and agentic workflows with improved efficiency and safety.
• GPT-5.4 Is Here: OpenAI’s Most Capable Model Yet Redefines Professional AI Work — GPT-5.4 integrates advanced reasoning, autonomous operation, and enterprise efficiency into a single AI system.
• GPT-5.4 Pro Hits 38% on FrontierMath, Why This Matters? — GPT 5.4 Pro integrates advanced reasoning, coding, and direct computer control, marking a potential turning point in AI capabilities.
• OpenAI just dropped GPT-5.5... (WOAH) — GPT 5.5 offers higher intelligence and token efficiency, especially for agentic coding and knowledge work, at a lower effective cost.
• OpenAI just dropped GPT-5.4 and WOW.... — GPT 5.4 unifies advanced coding, reasoning, and agentic capabilities into a single, highly efficient frontier model.
• OpenAI introduces GPT-5.4 with more knowledge-work capability — GPT-5.4 enhances agentic knowledge work through improved reasoning, context, visual understanding, and token efficiency.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.