aiaimodel.cards

GPT-5.4

Language modelPrevious generationCard summarized

A more affordable model for coding and professional work.

Active (legacy)● Re-checked 12 Oct 2026 · 104 facts · 0 corrections · 2 unconfirmedSource: GPT-5.4 Thinking System Card (38 pp)

Use it via API

$1,545 /month
$30.90 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

Supports

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

Good for

  • Complex professional work and coding (API docs: "a flagship model for complex professional work")
  • Computer use with platform and developer-configurable confirmation policies
  • Better reversion of its own changes while leaving user work intact on long agentic rollouts

Watch out for

Overview

OpenAI · 12GPT, front

GPT-5.4

A more affordable model for coding and professional work.

Cost
$2.50 in
Context
1.05M
Input / output
$2.50 / $15.00
Max output
128K
Knows up to
August 2025
Released
5 Mar 2026

OpenAI's system card for GPT-5.4 Thinking, the latest GPT-5 reasoning model at release. It is the first general-purpose OpenAI model with mitigations for High capability in cybersecurity, building on the approach for GPT-5.3-Codex in ChatGPT and the API. It is treated as High in biology and chemistry and in cybersecurity, and below High in AI self-improvement. Cyber results closely match GPT-5.3-Codex. p.4, p.16, p.20

gpt-5.4
Input / output per 1M
$2.50 / $15.00
Context window
1.05M
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. Every claim carries its source page.