01Overview02Models03Rankings04Compare05Quiz06Guide

Model

GPT-5.4 mini NO.14

Low-cost previous-generation OpenAI workhorse for standard routing.

CompareBack to the index

OpenAI · GPT · Fast · Closed weights

Specification

Fig.01 — GPT — OpenAI

USA

Context window
400Ktokens
Max output
128Ktokens
Input price
$0.75/ 1M
Output price
$4.50/ 1M
Cached input
$0.07/ 1M
Throughput
—
Class
Fast
Modalities
Text, Image, Pdf
Weights
Closed
Knowledge cutoff
Oct 2025
Released
Mar 2026

Overview

GPT-5.4 mini is OpenAI's affordable previous-generation workhorse, handling the bulk of standard routed traffic at a fraction of flagship cost. It delivers solid reasoning and coding for everyday tasks with a 400K context and full multimodal input. With GPT-5.6, OpenAI's fast tier is now led by the more capable Luna, but mini remains a cheaper option (with nano cheaper still) for cost-driven, high-volume routing.

Strengths

  • Strong value for standard production work
  • Good coding and reasoning for the price
  • 400K-token context and multimodal input
  • Fast and cheap enough for high volume
  • Steep prompt-cache discount

Trade-offs

  • Not frontier-reliable for long-horizon agentic coding
  • Below flagship on hard reasoning
  • Smaller context than the 5.4/5.5 flagships

Fit

Best for

  • Standard production routing
  • Cost-sensitive coding assistants
  • Chat and support backends
  • RAG and summarization at scale
  • Lightweight agents

Not ideal for

  • Hard autonomous coding runs
  • Frontier research-grade reasoning

Capabilities

Ten axes, normalized 0–100 and scored the same way across the whole catalog — so a 78 here means what a 78 means anywhere else on the bench.

Fig.02 — Capability radar

0 — 100

Reasoning
84
Coding
82
Math
85
Writing
85
Knowledge
84
Speed
85
Agentic
80
Vision
82
Multilingual
86
Long Context
85

Benchmarks

Public results, with independent third-party runs marked. Bars normalize percentages against 100 and Elo ratings against a 1500 ceiling.

Independent index

Artificial Analysis Intelligence Index

Composite of ~9–10 independent evals · Artificial Analysis

40/ 100

BenchmarkResult
Artificial Analysis Intelligence IndexIndependentGeneral · Artificial Analysis
40%
SWE-bench VerifiedIndependentCoding · SWE-bench
58.4%
MMLU-ProKnowledge · Vendor-reported
55.3%
DeepSWEIndependentCoding · Datacurve
24%

Alternatives

Models in roughly the same class — the ones worth weighing against this record.

Compare side by side
01GPT-5.6 LunaOpenAIThe fast, cost-efficient GPT-5.6 tier for everyday work at scale.02GPT-5.4 nanoOpenAIOpenAI's cheapest small model for high-volume simple work.03Gemini 3.5 FlashGoogle DeepMindFast, cheap multimodal workhorse with near-Pro quality.

Index

PreviousGPT-5.4NextGPT-5.4 nano
All models/EDUCATION