GPT-5.4 API

openai/gpt-5.4
by OpenAIrelease date: 3/5/2026

GPT-5.4 is OpenAI's most advanced model for complex professional tasks, offering agentic computer use, top-tier coding, and superior knowledge work abilities.

$1.25/$7.5per 1M tokens

GPT-5.4 API - Background

Overview

GPT-5.4 is OpenAI's latest frontier model, released in March 2026, and is positioned as the company's most powerful and efficient model for complex professional tasks. It represents a significant leap in agentic capabilities, native computer use, and unified coding plus reasoning abilities, making it highly suitable for advanced API-driven applications.

Development History

GPT-5.4 was officially launched on March 5, 2026, as the mainline successor to the GPT-5.2 and GPT-5.3-Codex models. Unlike previous incremental updates, GPT-5.4 fully integrates the advanced coding capabilities of Codex into the core model and introduces native computer control features. This marks a pivotal step in OpenAI's evolution towards agentic AI and robust knowledge work automation, with the GPT-5.4 API now serving as the primary interface for developers and enterprises.

Key Innovations

  • Native computer use capabilities, enabling direct control of computer interfaces and automation frameworks
  • Unified and enhanced coding abilities, surpassing previous Codex-level performance for end-to-end software development
  • Significantly improved knowledge work functions, including advanced document analysis, spreadsheet integration, and reduced error rates

GPT-5.4 API - Technical Specifications

Architecture

GPT-5.4 is based on a highly optimized transformer architecture, incorporating agentic planning modules and advanced tool-use integration. It supports multimodal inputs, extended context windows, and seamless orchestration of reasoning and code generation within the same API endpoint.

Parameters

While specific parameter counts are not disclosed, GPT-5.4 operates at a scale exceeding previous GPT-5.x models, supporting context windows up to 1.05 million tokens for API users, enabling complex, multi-step workflows.

Capabilities

  • Native computer interaction, including screen understanding and automated control via API
  • End-to-end software development, debugging, and architectural planning with human-level code quality
  • Advanced knowledge work, such as financial analysis, long document summarization, and cross-file reasoning

Limitations

  • High computational requirements for the most complex tasks, especially with extended context or deep reasoning
  • Some advanced tasks may experience latency, particularly on the Pro variant, requiring background processing

GPT-5.4 API - Performance

Strengths

  • Industry-leading performance in computer use benchmarks, outperforming human baselines
  • Exceptional consistency and quality in professional knowledge work and code generation

Real-world Effectiveness

Early user feedback and benchmark results confirm that GPT-5.4 API delivers substantial productivity gains for programmers, analysts, and knowledge workers. It achieves a 75% score on OSWorld Verified computer use benchmarks, surpassing human averages, and is recognized for producing outputs that closely match expert-level standards in document analysis, coding, and automation tasks.

GPT-5.4 API - When to Use

Scenarios

  • You have a need to automate complex workflows that involve interacting with desktop applications or web interfaces. The GPT-5.4 API is ideal for building agentic solutions that can understand screen content, plan actions, and execute mouse and keyboard operations, resulting in significant efficiency gains for IT support, RPA, and digital assistants.
  • You are developing large-scale software projects requiring advanced code generation, debugging, and architectural planning. The GPT-5.4 API integrates Codex-level coding abilities directly into the main model, enabling end-to-end project delivery, rapid prototyping, and seamless code review, which accelerates development cycles and improves code quality.
  • You need to process and analyze extensive business documents, financial reports, or presentations across multiple formats. The GPT-5.4 API excels at handling long-context inputs, performing cross-file analysis, and generating accurate summaries or insights, making it invaluable for financial analysts, consultants, and enterprise knowledge workers.

Best Practices

  • Leverage the GPT-5.4 API's upfront planning and interruptible reasoning features to guide outputs and reduce iteration cycles.
  • Utilize the model's extended context capabilities for tasks involving large documents or multi-step workflows to maximize accuracy and coherence.

Technical Specs

Context Length1,050,000
Release Date3/5/2026
Input Formats
textimage
Output Formats
textcodejson

Capabilities & Features

Capabilities
advanced reasoninglong context understanding (up to 1.05M tokens)native code generation (industry level, full-project)computer use/control via screen and automationtool use and plugin integrationknowledge work (document, financial, research tasks)multimodal input (text, image, limited audio)interruptible/plannable thinking processsecure computation (enhanced cybersecurity features)
Supported File Types
.txt.pdf.docx.xlsx.pptx.csv.jpg.png
GPT-5.4 API - Cheap API - OpenAI - Defapi