← Models

GPT-5.4

OpenAI · GPT-5

Current GPT-5.4 premium API model for difficult professional work, tool use, and computer-assisted tasks.

Part of GPT family · Other versions: GPT-5.6 Luna , GPT-5.6 Terra , GPT-5.6 Sol , GPT-5.5 Pro , GPT-5.5 , GPT-Rosalind , GPT-5.4 mini , GPT-5.4 nano , GPT-5.4 Pro , GPT-5.3 , GPT-5.3-Codex , GPT-5.2-Codex , GPT-5.2 Pro , GPT-5.2 , GPT-5-Codex , GPT-5 mini , GPT-5 nano , GPT-5 , GPT-4.1 , GPT-4o mini Transcribe , GPT-4o mini TTS , GPT-4o Transcribe , Computer Use Preview , GPT-4o mini , GPT-4o
Type
language
Context
1M tokens
Max Output
128K tokens
Status
current
Input
$2.5/1M tok
Output
$15/1M tok
API Access
Yes
License
proprietary
reasoning coding agentic tool-use long-context computer-use professional-workflows
Released March 2026 · Updated June 8, 2026

Overview

Freshness note: Model capabilities, limits, and pricing can change quickly. This profile is a point-in-time snapshot last verified on June 8, 2026.

GPT-5.4 is OpenAI’s still-current premium GPT-5.4 model for difficult professional work. OpenAI’s March 5 launch positioned it as a general-purpose top tier for harder coding, analysis, and agentic workflows, with native computer-use support and long-context handling across API and Codex-style execution surfaces. As of June 8, 2026, GPT-5.5 is the newest OpenAI frontier route across eligible ChatGPT, Codex, and API workflows, but GPT-5.4 remains relevant for existing integrations, compatibility, and cost-routing decisions.

Capabilities

GPT-5.4 is built for tasks where a mediocre answer is expensive: multi-step technical analysis, code review, decision-ready memos, and longer tool-using agent loops. The launch framing also matters operationally: OpenAI now treats computer use as part of the top-tier model story rather than a separate novelty surface. That makes GPT-5.4 more relevant for agents that need to inspect interfaces, reason over screenshots, or work through multi-step task flows.

Technical Details

OpenAI’s current model docs list GPT-5.4 with a 1,050,000-token context window, 128,000-token maximum output, text and image input, and text output. That is the value stored in this repo’s frontmatter. The same docs list streaming, function calling, structured outputs, web search, file search, image generation, code interpreter, hosted shell, apply patch, skills, computer use, MCP, and tool search support through the Responses API.

OpenAI also describes GPT-5.4 as the first mainline reasoning model to incorporate the frontier coding capabilities of GPT-5.3-Codex, while adding native computer use and tool-search improvements. That made it the March frontier general model, but not a guarantee that every Codex client defaults directly to GPT-5.4.

Pricing & Access

Current published standard API pricing is:

  • Input: $2.50 per 1M tokens
  • Cached input: $0.25 per 1M tokens
  • Output: $15.00 per 1M tokens

For GPT-5.4 and GPT-5.4 Pro, OpenAI also says prompts above 272K input tokens are priced at 2x input and 1.5x output for the full session, and regional-processing endpoints carry a 10% uplift. Teams evaluating the newest OpenAI quality tier should compare GPT-5.4 against the API-listed GPT-5.5 route rather than assuming GPT-5.4 is still the newest premium default.

Best Use Cases

Use GPT-5.4 for production assistants, code-heavy investigations, review workflows, and long-running tool-use chains where better reasoning quality justifies a premium over smaller GPT routes and an existing GPT-5.4 integration is already the right operational fit. It is still sensible when the workflow may escalate from drafting into analysis plus action instead of staying in pure chat.

Comparisons

  • GPT-5.5 (OpenAI): Newer frontier route across eligible ChatGPT, Codex, and API workflows.
  • GPT-5.2 (OpenAI): Older but still useful route when you want compatibility with existing GPT-5.2 integrations.
  • GPT-5.3-Codex (OpenAI): Important coding-specialized reference point that GPT-5.4 incorporates, but still not the same product role.
  • Claude Opus family (Anthropic): Another premium lane for complex reasoning and coding workloads.