← All models

GPT-5.4

Published Publisher OpenAIgpt-5.4

GPT-5.4

Release and positioning

Released on March 5, 2026, GPT-5.4 is OpenAI’s general-purpose reasoning model for complex professional work, software development, and multi-step agent workflows. The API alias is gpt-5.4; the dated snapshot is gpt-5.4-2026-03-05. (openai.com)

Capabilities and uses

  • Combines reasoning, coding, document understanding, image perception, web research, and tool use. Example applications include software engineering, financial and spreadsheet analysis, document preparation, and research across multiple sources.
  • Supports native computer use, allowing an application to build agents that inspect screenshots and interact with software through a user interface. Tool search can load relevant tool definitions on demand, and context compaction supports longer-running workflows.
  • In the API, it supports text and image input and text output, as well as function calling, structured outputs, and streaming. Reasoning effort can be set to none, low, medium, high, or xhigh; the default is none.
  • The documented API context window is 1,050,000 tokens, with up to 128,000 output tokens. Its knowledge cutoff is August 31, 2025. (developers.openai.com)

Limitations

The model can still produce factual errors; important outputs should be checked. Audio and video input are not supported by this model, and fine-tuning is not supported. Computer-use workflows should run in an isolated environment, with human oversight for high-impact actions. Context limits and availability can differ by product or platform. (openai.com)

OpenAI API pricing

For standard text-token usage, the official rate is USD $2.50 per million input tokens, USD $0.25 per million cached input tokens, and USD $15.00 per million output tokens. These are OpenAI API rates, not ChatGPT subscription prices or third-party platform prices. (openai.com)

For prompts exceeding 272,000 input tokens, standard, Batch, and Flex usage is charged at twice the applicable input rate and 1.5 times the applicable output rate for the full session. Regional data-residency processing adds 10%. OpenAI’s launch announcement states that Batch and Flex are priced at half the standard API rate; those discounts and the long-context and regional modifiers have their own conditions. Tool calls may incur separate charges. (developers.openai.com)

Pricing sources: OpenAI model documentation and official launch announcement. Pricing checked November 16, 2026.

Official sources