GPT 5.4 Pro

Supported Functionality

Description

GPT-5.4 Pro is OpenAI's compute-intensive professional reasoning model, released on March 5, 2026 as the Pro version of GPT-5.4 for difficult problems that benefit from smarter and more precise responses.

Compared with the standard version, it applies more reasoning compute to improve consistency on complex tasks. It supports medium, high, and xhigh reasoning effort, a 1,050,000-token context, and up to 128,000 output tokens. The model is available only through the Responses API, and requests can take several minutes, so OpenAI recommends background mode. It supports function calling and several retrieval and interaction tools but not Structured Outputs, Code Interpreter, Hosted Shell, Skills, or fine-tuning.

Key Capabilities

• Compute-Intensive Deep Reasoning: Uses additional test-time compute for difficult problems, prioritizing accuracy and consistency on complex professional tasks.
• Three Reasoning Levels: medium, high, and xhigh let developers scale reasoning effort according to task difficulty and value, with medium as the default.
• Long-Context Analysis: A 1,050,000-token context can hold large source collections, long documents, and complex task history, with up to 128,000 output tokens.
• Professional Analysis and Review: Suits research, technical review, and high-value decision support that require rigorous reasoning, evidence synthesis, and constraint balancing.
• Vision and Interface Understanding: Accepts image input and can use Computer Use to understand and interact with screenshot-based interfaces.
• Retrieval-Augmented Reasoning: Web Search and File Search can incorporate online information and file evidence into complex analysis.
• Tool and System Connectivity: Function calling, MCP, Tool Search, Apply Patch, and image generation can connect external systems and perform controlled actions.

Technical Strengths