OpenAI logoOpenAIopenai.com

GPT-5.4-mini

GPT-5.4-mini is OpenAI's high-volume workhorse — the model you reach for when you run millions of requests and cost per call decides the project.

Context window
400,000 tokens
Max output
128,000 tokens
Knowledge cutoff
Aug 2025
API identifier
gpt-5.4-mini

Pricing · per 1M tokens

Input$0.75
Output$4.50
Cached input$0.07

Cached input is ~10× cheaper — it pays off when you send the same context again and again.

Best for

Code GenerationCustomer Support AgentsData Extraction & RAG

Industries

E-commerceFinanceRetail

Business functions

Information TechnologyCustomer ServiceFinance & Accounting

Overview

What GPT-5.4-mini is good at

OpenAI calls it their strongest mini model yet for coding, computer use, and subagents. In plain terms, you get much of GPT-5.4's capability at faster speed and lower cost per call. That trade is the whole point.

Use it where volume is the problem. Support agents answering thousands of tickets. Pipelines pulling data from documents all day.

Coding agents running across your whole repo, too. When one request costs very little, running it a million times stays affordable.

It has room to work. The context window holds 400,000 tokens, and it can write up to 128,000 tokens at once. So long documents and big code files fit without splitting them up.

Be honest about the job, though. For your hardest reasoning — the tricky architecture call, the gnarly multi-step problem — step up to full GPT-5.4. Mini is tuned for throughput, not the deepest thinking.

The knowledge cutoff is August 31, 2025. Reach for GPT-5.4-mini when you're running many requests and cost per call is what matters most.

Real numbers

What GPT-5.4-mini costs in practice

Real enterprise workloads, with the token assumptions shown openly. Your mileage varies — these are honest starting points, not guesses.

RetailCustomer Service

Support ticket triage and reply drafting

$0.0022per request
$675per month

A large retailer gets thousands of support tickets a day about orders, returns, and sizing. GPT-5.4-mini reads each ticket, pulls the order context, and drafts a reply your agent can send in one click. At this volume, a pricier model would blow the budget — the low cost per ticket is what makes full coverage possible.

1,500 input tokens250 output tokens300,000 requests / month

Your system prompt and product docs repeat on every ticket, so caching that input keeps the cost per reply low. Output tokens drive most of the bill here.

FinanceFinance & Accounting

Invoice and statement data extraction

$0.0033per request
$2,640per month

A finance team processes hundreds of thousands of invoices and bank statements every month. GPT-5.4-mini reads each document and returns clean, structured data — vendor, amount, date, line items — ready for your books. Running this on a top-tier model would cost more than the work saves, so the low cost per document is the whole reason it pencils out.

2,000 input tokens400 output tokens800,000 requests / month

The extraction schema repeats on every document, so cached input carries most of the cost down. This job is input-heavy, so trimming documents to the pages you need helps.

E-commerceInformation Technology

Automated code review and test generation

$0.0099per request
$594per month

An e-commerce engineering team runs automated agents across every pull request — reviewing diffs, writing tests, and flagging bugs. GPT-5.4-mini handles the routine passes fast, so your engineers only step in for the hard calls. Because each run costs so little, you can afford to check every PR instead of a lucky few.

6,000 input tokens1,200 output tokens60,000 requests / month

Coding needs a lot of context, so cached repo files lower the input cost across runs. Output is where cost lands — and for the hardest refactors, escalating to GPT-5.4 is worth it.

How we work these out: cost = (input ÷ 1M × $0.75) + (output ÷ 1M × $4.50), then × monthly volume. List prices only, no cached-input discount applied — so these are the ceiling, not the floor.

FAQ

Common questions about GPT-5.4-mini

How much does GPT-5.4-mini cost?+

GPT-5.4-mini costs $0.75 per million input tokens and $4.50 per million output tokens. Cached input drops to $0.075 per million, a tenth of the standard input price. So if your prompts repeat — like a fixed system prompt across many calls — caching cuts your input cost sharply. Output tokens are the bigger line item, so shorter replies keep your bill down.

GPT-5.4-mini vs GPT-5.4: which should you use?+

Pick GPT-5.4-mini for volume, GPT-5.4 for depth. Mini gives you much of GPT-5.4's capability at faster speed and lower cost per call — ideal when you run thousands or millions of requests. For your hardest reasoning, like deep multi-step problems or tricky judgment calls, step up to full GPT-5.4. Many teams use both: mini for the bulk work, GPT-5.4 for the few calls that need it.

What is GPT-5.4-mini's context window?+

GPT-5.4-mini has a 400,000-token context window. It can also write up to 128,000 tokens in a single response. That's room for long documents, big code files, or a lot of chat history in one call — no chunking needed. As a rough feel, 400,000 tokens is around 300,000 words of input.

Is GPT-5.4-mini good for high-volume customer support?+

Yes — high-volume customer support is one of the jobs GPT-5.4-mini is built for. You can triage tickets, draft replies, and tag conversations across thousands of messages a day, at a low cost per message. Put your product docs or FAQ in a cached system prompt, and the repeated context costs you even less. For a rare, sensitive escalation that needs careful judgment, route just that ticket to GPT-5.4.

GPT-5.4-mini vs GPT-4o mini: what's the difference?+

GPT-5.4-mini is the newer generation — OpenAI's strongest mini yet for coding, computer use, and subagents. It brings a 400,000-token context window and an August 2025 knowledge cutoff. If you're choosing today for high-volume coding or agent work, GPT-5.4-mini is the one built for it. For GPT-4o mini's exact price and limits, check OpenAI's current docs, since those numbers can change.

Pricing as of 8 Jul 2026 · reviewed monthlySources: OpenAI API pricing · OpenAI model docs

Not sure GPT-5.4-mini fits your workload?

We'll size the right model for your use case and give you a real cost estimate. One free call, no slides.

Talk to us