Overview
GPT-5.4 Pro is the higher-compute version of GPT-5.4, designed for demanding tasks where users prioritize more precise and consistent answers over speed and cost. OpenAI announced it on March 5, 2026, for ChatGPT and the API. The API alias is gpt-5.4-pro; the dated snapshot is gpt-5.4-pro-2026-03-05. (openai.com)
Capabilities and uses
- Supports text and image inputs and text output. It can be used for complex reasoning, research, coding, and professional workflows involving documents, spreadsheets, and presentations.
- Supports function calling and, through the Responses API, tools including web search, file search, image generation, computer use, MCP, and tool search.
- The API model documentation specifies a 1,050,000-token context window, a maximum of 128,000 output tokens, and an August 31, 2025 knowledge cutoff. Supported reasoning effort settings are
medium(default),high, andxhigh. (developers.openai.com)
Limitations
The model can make factual or reasoning errors, so important outputs should be checked. Some requests can take several minutes; OpenAI recommends background mode to help avoid timeouts. The model is available through the Responses API rather than Chat Completions. The model documentation lists audio and video as unsupported, as well as structured outputs, fine-tuning, and Code Interpreter. Availability and rate limits depend on the user's API tier. (developers.openai.com)
API pricing
OpenAI's announced standard API text-token rates are USD $30 per million input tokens and USD $180 per million output tokens. The announcement does not list a cached-input price for GPT-5.4 Pro. These are API token charges, not a ChatGPT subscription price. (openai.com)
For prompts exceeding 272,000 input tokens, the model documentation specifies a 2× input and 1.5× output multiplier for the full session under standard, batch, and flex processing. Applied to the standard rates above, this is USD $60 per million input tokens and USD $270 per million output tokens. Regional-processing endpoints carry a 10% uplift. Tool use may incur additional charges. A model-specific batch or flex rate was not independently verified here. Pricing checked November 16, 2026. (developers.openai.com)