GPT 5.4

Supported Functionality

Description

GPT-5.4 is OpenAI's frontier general reasoning model for complex professional work, coding, and agentic workflows. OpenAI provides the gpt-5.4 alias and the dated gpt-5.4-2026-03-05 snapshot, indicating a March 5, 2026 version snapshot; its knowledge cutoff is August 31, 2025. It accepts text and image inputs, produces text, and offers a 1,050,000-token context window with up to 128,000 output tokens. OpenAI has not disclosed its parameter count or detailed architecture.

Its central advance is combining a million-token context, five reasoning-effort levels from the default none through xhigh, and a broad agent toolchain in one model. In the Responses API, it supports Web Search, File Search, image generation as a tool, Code Interpreter, Hosted Shell, Apply Patch, Skills, Computer Use, MCP, and Tool Search.

Key Capabilities

• Adjustable Frontier Reasoning: Five levels—none, low, medium, high, and xhigh—span low-latency answers through difficult professional analysis.
• Million-Token Long Context: A 1,050,000-token window accommodates large corpora, repositories, and extended sessions for cross-document synthesis and sustained project work.
• Professional Software Engineering: Hosted Shell, Code Interpreter, and Apply Patch support code analysis, editing, testing, and iterative verification.
• Agent Tool Orchestration: Function calling, Tool Search, Skills, and MCP discover capabilities across large tool surfaces and execute multi-step tasks.
• Vision and Interface Understanding: Reads screenshots, charts, document pages, and design references for visual analysis, extraction, and frontend implementation.
• Computer Interaction: Computer Use lets the model operate graphical interfaces in controlled environments for application workflows that lack direct APIs.
• Retrieval and Structured Output: Web Search, File Search, and Structured Outputs combine current or private information with machine-readable, verifiable results.

Technical Strengths