GPT-5.4 Launches with Native Computer Use and 1M Token Context

30-Second Summary

OpenAI released GPT-5.4 on 5 March 2026, its most capable general-purpose model to date. The release marks a significant shift: for the first time, a mainline OpenAI model ships with native computer-use capabilities, allowing agents to operate applications, click through interfaces, and execute multi-step workflows without custom integrations. Combined with a 1-million-token context window, improved factual accuracy, and a new Tool Search system for the API, GPT-5.4 represents a meaningful step toward practical autonomous task execution for business operators.

At a Glance

• Topic: Model Releases
• Company: OpenAI
• Date: 5 March 2026
• Announcement: Launch of GPT-5.4 across ChatGPT, the OpenAI API, and Codex
• What Changed: First mainline OpenAI model with native computer-use capabilities; 1M-token context window; 33% fewer false claims versus GPT-5.2
• Why It Matters: Autonomous task execution across applications is now accessible to any team using the OpenAI API, without needing to build custom tool integrations per application
• Who Should Care: Business operators running AI-assisted workflows, developers building automation tools, and enterprise teams evaluating AI agent platforms

Key Facts

• Company: OpenAI
• Launch Date: 5 March 2026
• What Changed: Native computer use, 1,050,000-token context window, 33% reduction in false claims, 18% reduction in error-containing responses, new Tool Search API feature
• Who It Affects: ChatGPT Plus, Team, Pro, Enterprise, and Edu subscribers; API developers; Codex users
• Primary Source: OpenAI official announcement, OpenAI API documentation, TechCrunch

What Happened

OpenAI released GPT-5.4 on 5 March 2026, describing it as the first general-purpose frontier model to combine state-of-the-art coding capabilities with native computer-use support. The release was simultaneous across ChatGPT, the OpenAI API, and Codex.

The most significant new capability is computer use. GPT-5.4 can now operate computers as an agent, reading screens and executing tasks across applications without requiring custom integrations for each tool. This makes it possible to build agents that handle multi-step workflows across different software, including tools that have no API. The model supports up to 1,050,000 tokens of context, enabling agents to plan, execute, and verify tasks across long workflows without losing earlier context.

On accuracy, OpenAI reports that GPT-5.4's individual claims are 33% less likely to be false than those of GPT-5.2, and full responses are 18% less likely to contain any errors. A new Tool Search system for the API changes how tool definitions are handled: instead of loading all tool definitions into the system prompt at the start of each request, the model looks up tools as needed. This reduces token usage and cost in systems with many available tools.

GPT-5.4 is available in three variants: the standard model, GPT-5.4 Thinking (a reasoning-optimised version replacing GPT-5.2 Thinking for Plus, Team, and Pro users), and GPT-5.4 Pro (available to Pro and Enterprise plans). Enterprise customers can enable early access through admin settings. API pricing starts at $2.50 per million input tokens and $15.00 per million output tokens. The Batch API option reduces costs by 50% for asynchronous jobs.

Why It Matters

• Computer use as a native capability removes a major barrier to building autonomous agents. Previously, agents needed custom integrations or browser automation libraries to interact with applications. GPT-5.4 handles this natively.
• The 33% reduction in false claims and 18% reduction in error-containing responses materially improves the reliability of AI-generated content in business workflows, reducing the cost of review and correction.
• The 1-million-token context window enables agents to work across entire document sets, code repositories, or conversation histories in a single session, without truncating or chunking data.
• Tool Search reduces API costs in complex agentic systems by loading tool definitions on demand rather than front-loading them all into each request.
• Enterprise-grade infrastructure, including Zero Data Retention and regional data residency endpoints, means GPT-5.4 can be deployed in compliance-sensitive environments.
• GPT-5.2 Thinking is retiring on 5 June 2026, creating a migration deadline for teams currently using it.

The David and Goliath View

GPT-5.4 is the clearest signal yet that the frontier of AI capability is no longer about language. It is about action. A model that can read a screen, click a button, fill a form, and move between applications is not a better chatbot. It is the foundation of a digital worker.

For operators running lean teams, this is consequential. The traditional barrier to automation was integration: every tool you wanted to automate required its own API connection, its own custom code, and its own maintenance overhead. Computer use sidesteps that entirely. If a human can do it on a screen, an agent built on GPT-5.4 can, in principle, do it too.

The practical implication is this: if your organisation has been waiting for AI to handle real tasks rather than just answer questions, the technical foundation is now in place. The constraint has shifted from model capability to workflow design and governance. Start by identifying two or three repetitive, screen-based tasks your team performs daily. Those are your first automation candidates.

Where This Fits in the AI Stack

AI Growth Engine: Computer-use agents can handle outbound research, CRM updates, lead qualification workflows, and report generation across tools that have no API, dramatically expanding what can be automated in a revenue operation.

Employee Amplification Systems: GPT-5.4's improved accuracy and long context make it a stronger foundation for internal knowledge tools, document analysis, and cross-application task automation, reducing the manual overhead on knowledge workers.

Secure AI Brain: Enterprise Zero Data Retention and regional data residency endpoints mean GPT-5.4 can be deployed in compliance-sensitive environments. Organisations handling sensitive data can use the full capability set without routing data through shared infrastructure.

Questions Operators Are Asking

What does "computer use" actually mean for my business? It means an AI agent can interact with software through its user interface, the same way a human would. If your team uses a tool that has no API, such as a legacy system or a heavily customised internal platform, a computer-use agent can still automate tasks in it. The limitation is speed and reliability compared to direct API calls, but the coverage is far broader.

Is the 1-million-token context window available right now? Yes, but it requires explicit configuration. By default, GPT-5.4 uses a 272K-token context window. You need to set model_context_window and model_auto_compact_token_limit parameters to access the full 1,050,000-token window. Note that prompts exceeding 272K tokens are charged at 2x the standard input rate.

How much better is GPT-5.4 accuracy in practice? OpenAI reports 33% fewer false claims at the individual claim level and 18% fewer responses containing any errors, compared to GPT-5.2. For business outputs where accuracy matters, such as financial summaries, legal drafts, or client-facing reports, this is a meaningful improvement that reduces the cost of human review.

We are still on GPT-5.2 Thinking. Do we need to migrate? Yes, and you have until 5 June 2026. GPT-5.2 Thinking will remain available in the Legacy Models section for paid users until that date, after which it will be retired. GPT-5.4 Thinking is the direct replacement and is available now.

What is Tool Search and should we use it? Tool Search is a new API feature that loads tool definitions on demand rather than including them all in the system prompt from the start. If you are building agentic systems with many tools, this reduces token consumption and cost per request. It is worth enabling for any workflow that uses more than five or six tools.

Citable Summary

What happened: OpenAI launched GPT-5.4 on 5 March 2026, its most capable general-purpose model to date, featuring native computer-use capabilities, a 1-million-token context window, 33% fewer false claims than GPT-5.2, and a new Tool Search system for the API.

Why it matters: Computer use as a native capability removes the integration barrier for AI automation, enabling agents to operate any software a human can use. Combined with improved accuracy and long context, GPT-5.4 makes autonomous workflow execution practical for business operators without requiring custom integrations per tool.

David and Goliath view: The frontier has shifted from language to action. Operators with lean teams should identify repetitive, screen-based workflows as immediate automation candidates, then build governance frameworks before scaling agent deployments.

Offer relevance:

• AI Growth Engine: computer-use agents for revenue workflows across tools with no API
• Employee Amplification Systems: improved accuracy and long context for internal knowledge and task automation
• Secure AI Brain: enterprise Zero Data Retention and data residency for compliance-sensitive deployments