Overview
GPT-5.4 mini is a faster, more efficient GPT-5.4-family model designed for high-volume workloads, especially coding, computer use, and subagents. It was released on March 17, 2026. The API alias is gpt-5.4-mini; the versioned snapshot is gpt-5.4-mini-2026-03-17. (openai.com)
Capabilities and uses
- Accepts text and image input and produces text output. It supports reasoning, streaming, function calling, and structured outputs. Reasoning effort can be set to
none,low,medium,high, orxhigh; the default isnone. (developers.openai.com) - Supports tools including web search, file search, computer use, code interpreter, and tool search through the Responses API. Suitable uses include coding assistants, debugging, codebase navigation, screenshot interpretation, and delegated subagent tasks. (developers.openai.com)
- OpenAI reports that the model runs more than twice as fast as GPT-5 mini. In its evaluations at
xhighreasoning effort, GPT-5.4 mini scored 54.4% on SWE-Bench Pro and 72.1% on OSWorld-Verified. Actual latency varies by workload. (openai.com)
Limits
- The API context window is 400,000 tokens, with a maximum of 128,000 output tokens. The knowledge cutoff is August 31, 2025. (developers.openai.com)
- Audio and video are not supported as native modalities, and fine-tuning is not supported. Image input should not be confused with native image output; the model can use a separate image-generation tool. (developers.openai.com)
- Long-context retrieval is not perfect: OpenAI reported 33.6% on its MRCR v2 8-needle evaluation at 128K–256K context. Rate limits vary by API usage tier. (openai.com)
Pricing
For OpenAI API text tokens, standard pricing is in US dollars (USD): $0.75 per 1 million input tokens, $0.075 per 1 million cached input tokens, and $4.50 per 1 million output tokens. Cached-input pricing applies to eligible cached tokens. Regional processing endpoints for data residency carry a 10% uplift; tool usage may incur additional charges. These are API token rates, not Codex or ChatGPT subscription prices. Pricing checked November 16, 2026. (developers.openai.com)