aiaimodel.cards

GPT-5.4 Mini

Language modelPrevious generationCard summarized

Our strongest mini model yet for coding, computer use, and subagents

Active (legacy)● Re-checked 12 Oct 2026 · 58 facts · 0 correctionsSource: GPT-5.4 Thinking System Card, Appendix: GPT-5.4 mini (38 pp)

Use it via API

$464 /month
$9.27 per 1K requests
RAG over docs: 12K in · 600 out · 50K requests/mo · 30% cached · change
Code exampleCompare

Where to call it

  • OpenAI APIgpt-5.4-mini

Supports

  • Image input
  • Streaming
  • Structured outputs
  • Reasoning control
  • Tool calling
  • Web search
  • Prompt caching
  • Batch API
Full feature support

Good for

  • Coding, computer use and subagents ("Our strongest mini model yet")
  • High-volume workloads that need GPT-5.4 strengths at lower latency and cost

Watch out for

Overview

OpenAI · 11mini, front

GPT-5.4 mini

Our strongest mini model yet for coding, computer use, and subagents

Cost
$0.75 in
Context
400K
Input / output
$0.75 / $4.50
Max output
128K
Knows up to
August 2025
Released
17 Mar 2026

A short appendix to the GPT-5.4 Thinking system card, added on 17 March 2026. OpenAI rates GPT-5.4 mini below High in biology and chemistry, cybersecurity and AI self-improvement. Its biology capability is closely comparable to GPT-5, which OpenAI now judges was not High, yet it received full safety training, including for biological risk. It has the lowest chain-of-thought controllability of any model OpenAI has reported. p.33–36

gpt-5.4-mini
Input / output per 1M
$0.75 / $4.50
Context window
400K
Max output
128K
Knows up to
August 2025
Reasoning
Configurable (effort none → xhigh)
Weights
Closed

About this card. A model card is the maker's own account of a model: what it's for, how it was trained and tested, and where it falls short. This page condenses OpenAI's GPT-5.4 Thinking System Card, Appendix: GPT-5.4 mini. Every claim carries its source page.