AI modelsOpenAI

GPT-5.4 nano

Decision matrix

GPT-5.4 nano model decision matrix
Overview

GPT-5 mini is OpenAI's smaller, cheaper sibling to GPT-5, priced at $0.25 per million input tokens and $2 per million output tokens, roughly a fifth of GPT-5's price. Pick it as the default for routine, high-volume API tasks; step up to GPT-5 only when a task genuinely needs more capability.

Other good fits
  • A cost-sensitive application where per-call price matters and the task doesn't demand the flagship model.
Pros
Read pros

GPT-5 mini is OpenAI's smaller, faster API model, built for routine tasks at a fraction of GPT-5's per-token price. It's the model to default to for high-volume, well-scoped work, classification, extraction, short drafts, where the flagship model's extra capability wouldn't change the outcome but would meaningfully increase the bill.

Pick it when…

  • You're making a lot of similar, simple calls (tagging, routing, short summarisation) where cost per call adds up.
  • The task is well-scoped enough that a smaller model reliably gets it right.
  • You want to prototype an application cheaply before deciding whether any task actually needs the flagship model.
Cons
Read cons

A smaller, cheaper model will strain on tasks that need real reasoning or careful judgment; don't assume it can substitute for GPT-5 on every task just because it's cheaper. Test on a representative sample before committing a production workload to it, since a model that looks fine on a handful of examples can still fail more often than expected at scale.

Pick something else when…

  • The task needs deeper reasoning or handles higher-stakes output. GPT-5 is the reliable step up.
  • You want to chat interactively, not build against an API. ChatGPT is the consumer product for that.
  • You want to compare GPT-5 mini against similarly priced models from other vendors. OpenRouter makes that comparison easy from one integration.

Avoid if

  • You just want to chat, not build against an API; [ChatGPT](tool page) is the consumer product for that.
Context

Not documented yet.

Requirements

Not documented yet.

Licence & cost

Prices checked 23 September 2026 (pricing (opens in a new tab)): $0.25 per million input tokens, $2 per million output tokens, about a fifth of GPT-5's rate on both counts. There's no subscription; billing is per token through your OpenAI API account. At high call volume, even this low per-token rate adds up, so it's worth tracking total spend across a busy application, not just the per-call price.

First task

Documentation-based suggestion — not run here.

Build a simple classifier: feed it fifty made-up support messages and ask for a one-word category label on each, then check its labels against ones a person assigned to the same messages.

Done when GPT-5 mini agrees with the human labels on most of the fifty, and you've identified the specific pattern behind any disagreement, so you know whether GPT-5 mini is reliable enough for this task or whether it needs the flagship model instead.

Links & evidence
Documentation-based · not hands-on · Sep 23, 2026

Rests on OpenAI's API pricing page, read on 2026-09-23. No API call was made for this review; context-window size was not published on the pricing page and should be confirmed on the model's own documentation page.

  • OpenAI API pricing · Sep 23, 2026 — GPT-5 mini at $0.25/$2 per million input/output tokens versus GPT-5 at $1.25/$10.
  • OpenRouter pricing · Sep 23, 2026 — Comparison alternative for testing cheap models from several vendors.
  • ChatGPT pricing · Sep 23, 2026 — Comparison alternative for chatting directly rather than building against the API.
Further review notes & alternatives

Connection

Call GPT-5 mini through the standard OpenAI API, specifying its model name in place of GPT-5's. Because both follow the same API conventions, switching between them for a comparison is typically a one-line change.

Alternatives and trade-offs

  • GPT-5 wins the moment a task needs real reasoning, not just fast, cheap pattern-matching.
  • Claude Haiku wins as the closest cross-vendor comparison for cheap, high-volume calls; worth benchmarking both on your actual workload.
  • OpenRouter wins when you want to try several cheap models from different vendors on the same task.

Learn more

Catalogue facts

Classify, extract, route, and draft short responses at high volume. Connect simple calls to OpenAI tools and integrations. Choose it when tasks do not require the hardest reasoning.

OpenAI · paid · cloud

  • high volume
  • text classification
  • entity extraction
  • request routing
  • short draft
  • large context
  • image input
  • eu residency
  • automate
  • draft-docs
  • write-code
  • talk-think

Browse all tools