GPT-5.4

GPT-5.4 is OpenAI's most capable frontier model as of March 2026, described by the company as its most capable and efficient model for professional work. It ships in three variants: GPT-5.4 (standard), GPT-5.4 Thinking (a reasoning-focused variant that presents initial reasoning plans before generating full answers), and GPT-5.4 Pro (optimized for enterprise and high-performance tasks).

Context and Efficiency

The model supports a 1 million token context window through the API (922K input, 128K output), the largest OpenAI has offered. GPT-5.4 uses up to 47% fewer tokens than GPT-5.2 on some tasks, translating to lower costs and faster responses.

Native Computer Use

GPT-5.4 is the first general-purpose GPT model with native computer-use capabilities, enabling agents to interact directly with software environments — navigating applications, executing workflows, and completing multi-step tasks across tools. It achieves record results on OSWorld-Verified and WebArena Verified benchmarks for computer use.

Tool Search

The API introduces Tool Search, a new system where the model looks up tool definitions on demand rather than loading all definitions upfront in the system prompt. This reduces token consumption and latency in agentic systems with large tool ecosystems.

Benchmarks

On GDPval, which tests agents across 44 professional occupations, GPT-5.4 matches or exceeds industry professionals in 83.0% of comparisons, up from 70.9% for GPT-5.2. Individual claims are 33% less likely to be false, and full responses are 18% less likely to contain errors compared to its predecessor. It also achieves top performance on the APEX-Agents benchmark for professional domain tasks.

Safety

OpenAI introduced chain-of-thought reasoning safety evaluation for GPT-5.4. The Thinking variant shows lower deception risk compared to previous reasoning models.

Availability

GPT-5.4 is available to ChatGPT Plus, Team, and Pro subscribers, as well as through the OpenAI API. The Pro variant is restricted to Pro and Enterprise plans. The release positions GPT-5.4 as a direct competitor to Claude Opus 4.6 and Gemini 3.1 Pro in the frontier model tier.