NEWHow AI models draw the world map · Gecko Tests
Beta
Home/Models/GPT-5.4 Mini
OpenAI logo

GPT-5.4 Mini

by OpenAI · Released Mar 2026

Multimodal
35.9
avg score
Rank #220
Compare
Better than 29% of all models
Context
400K tokens (~200 books)
Input $/1M
$0.75
Output $/1M
$4.50
Type
multimodal
License
Proprietary
Benchmarks
29 tested
Data as of
About

GPT-5.4 mini brings the core capabilities of GPT-5.4 to a faster, more efficient model optimized for high-throughput workloads. It supports text and image inputs with strong performance across reasoning, coding,...

Tested on 29 benchmarks · BenchGecko score 35.9. Top scores: OTIS Mock AIME 2024-2025 (88.9%), GPQA diamond (82.5%), LiveBench — Coding (74.7%).

Looking for similar performance at lower cost?
Qwen3 8B scores 36.3 (101% as good) at $0.12/1M input · 84% cheaper
Capabilities
coding
44.8
#119 globally
reasoning
38.0
#105 globally
math
36.2
#170 globally
knowledge
41.0
#199 globally
agentic
24.6
#43 globally
general
34.4
#109 globally
language
30.3
#145 globally
speed
42.1
#54 globally
Benchmark Scores
Compare All
Tested on 29 benchmarks · Ranked across 8 categories
Score Distribution (all 312 models)
0255075100
▲ You are here
codingCompare coding →
LiveBench — Coding

Regularly refreshed coding problems that avoid data contamination. New problems added monthly to prevent memorization.

74.7·
WeirdML

Unusual and adversarial machine learning challenges. Tests robustness of reasoning about edge cases in ML systems.

60.3·
Frontiercode
27.0·
reasoningCompare reasoning →
ARC-AGI

Abstraction and Reasoning Corpus. Tests fluid intelligence through novel visual pattern recognition puzzles. Core measure of general intelligence.

63.7·
LiveBench — Data Analysis

Fresh data analysis tasks testing ability to interpret tables, charts, and statistical data.

47.4·
LiveBench — Reasoning

Regularly refreshed reasoning problems testing logical deduction, spatial reasoning, and analytical thinking.

21.9·
mathCompare math →
OTIS Mock AIME 2024-2025

Mock AIME (American Invitational Mathematics Exam) problems from OTIS. Tests mathematical competition performance.

88.9·
FrontierMath-Tiers-1-3-v2-Private
51.2·
LiveBench — Mathematics

Regularly updated math problems that test numerical reasoning, algebra, calculus, and combinatorics.

37.0·
Quick compare:Anthropic logovs Claude Haiku 4.5Mistral AI logovs Mistral Large 2411Alibaba Qwen logovs Qwen2.5-Max
Excellent (85+) Good (70-85) Average (50-70) Below (<50)
Model Family · OpenAI GPT-5.4
Mar 2026
66.0
$2.50/M in1.1Mctx34 benchmarks
Mar 2026
$1.25/M in(-1.25)1.1Mctx
Apr 2026
$8.00/M in(+6.75)272Kctx(-778K)
Mar 2026
35.9-30.1
$0.75/M in(-7.25)400Kctx(+128K)29 benchmarks
Mar 2026
$0.38/M in(-0.38)400Kctx
Mar 2026
28.1-7.8
$0.20/M in(-0.17)400Kctx27 benchmarks
Mar 2026
$0.10/M in(-0.10)400Kctx
Mar 2026
80.7+52.6
$30.00/M in(+29.90)1.1Mctx(+650K)12 benchmarks
Mar 2026
$15.00/M in(-15)1.1Mctx
See the full GPT-5.4 family →
Similar Models
Anthropic logo
Claude Haiku 4.5
Anthropic
35.9$1.00/1M
Mistral AI logo
Mistral Large 2411
Mistral AI
35.8$2.00/1M
Alibaba Qwen logo
Qwen2.5-Max
Alibaba Qwen
36.2n/a
Links
Info
OpenAIPricing explorerDevelopers · API
Research
Technical Report
Documentation
API DocsPlayground
Community
@OpenAI
BenchGecko API
gpt-5-4-mini
Specifications
  • Typemultimodal
  • Context400K tokens (~200 books)
  • ReleasedMar 2026
  • LicenseProprietary
  • StatusActive
  • Cost / Message~$0.006
Available On
OpenAI logoOpenAI$0.75
Categories
codingreasoningmathknowledgeagenticgenerallanguagespeed
Learn More
context-windowtransformertokens
Share & Export
Tweet
Related Models
Anthropic logoClaude Haiku 4.5
35.9
Mistral AI logoMistral Large 2411
35.8
Alibaba Qwen logoQwen2.5-Max
36.2
Alibaba Qwen logoQwen3 8B
36.3
GPT-5.4 Mini is a proprietary multimodal AI model by OpenAI, released in March 2026. It has an average benchmark score of 35.9. Context window: 400K tokens.

Related Models

Claude Haiku 4.5 · AnthropicMistral Large 2411 · Mistral AIQwen2.5-Max · Alibaba QwenQwen3 8B · Alibaba QwenLlama 3.2 90B · Meta

Benchmarks

OTIS Mock AIME 2024-2025GPQA diamondLiveBench — CodingDtbenchARC-AGI

Related Pages

OpenAI · ProviderOpenAI · EconomyAll ModelsCompare ModelsPricingDevelopers · APIContext Window · Glossary

Key facts · as of 2026-10-05

  • GPT-5.4 Mini by OpenAI. BenchGecko score 35.9, rank 219 of 312 scored models (normalized average of public benchmark scores).
  • List price $0.75 input · $4.50 output per 1M tokens (as of 2026-10-05).
  • Sold by 5 providers (as of 2026-10-05): OpenAI $0.38 in / $2.25 out · Azure $0.75 in / $4.50 out · OpenAI $0.75 in / $4.50 out · Azure $0.82 in / $4.95 out · OpenAI $1.50 in / $9.00 out. Every provider

How to cite · data as of 2026-10-05

GPT-5.4 Mini · benchmarks, pricing and providers. BenchGecko, data as of 2026-10-05. https://benchgecko.ai/model/gpt-5-4-mini

Credit "Source: BenchGecko" with a link. Prices per provider and Gecko Tests are BenchGecko data (CC BY 4.0); benchmark scores keep their original source, listed in the JSON. JSON · llms.txt · MCP