supermarket

Vision-language

GPT-5.4 mini

By OpenAI, in the GPT-5 family.

Released
17 Mar 2026
Status
Legacy
Context
400K
Max output
128K
Input per 1M tokens
$0.75
Output per 1M tokens
$4.50
Open weights
No
License
proprietary
Parameters
—
Knowledge cutoff
31 Aug 2025
Reasoning
Yes
Modalities
text, image → text

Category scores

Out of 100, from independent and third-party results only, by method category-2026-09-29.1.

Category scores
CategoryScoreRankCoverage
Knowledge & reasoning34.0452 of 4
Coding & software engineering27.0Not ranked1 of 5
Agentic tool use36.3Not ranked1 of 2
Factuality & hallucination29.4541 of 1
Human preference45.4383 of 5

Benchmark results

Every result we hold for this model, with who ran it and where it was published.

Knowledge & reasoning

Knowledge & reasoning results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Chess PuzzlesPuzzles solved24.0%Independentxhigh7 Aug 2026Epoch AI
Chess PuzzlesPuzzles solved18.0%Independenthigh15 Apr 2026Epoch AI
Chess PuzzlesPuzzles solved3.0%Independentnone7 Aug 2026Epoch AI
DTBench (Decision Theory Benchmark)Accuracy (capability questions)80.0%Third partyxhigh—https://conceptualreasoning.ai/dtbench via Epoch AI
LMCA (Language Model Conceptual Argumentation)Correlation with expert ratings (Pearson r)0.4079Third partyxhigh—https://conceptualreasoning.ai/lmca via Epoch AI
Mystery Game PuzzlesPuzzles solved11.0%Independentnone27 Aug 2026Epoch AI
Mystery Game PuzzlesPuzzles solved8.0%Independentlow27 Aug 2026Epoch AI
Mystery Game PuzzlesPuzzles solved7.0%Independentmedium27 Aug 2026Epoch AI

Scientific reasoning

Scientific reasoning results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
GPQA DiamondAccuracy86.9%Independentxhigh7 Aug 2026Epoch AI
GPQA DiamondAccuracy83.6%Independenthigh15 Apr 2026Epoch AI
GPQA DiamondAccuracy64.1%Independentnone7 Aug 2026Epoch AI

Mathematics

Mathematics results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
FrontierMath Tier 4Accuracy9.8%Independentxhigh12 Jun 2026Epoch AI
FrontierMath Tier 4 (v1)Accuracy2.1%Independenthigh15 Apr 2026Epoch AI
FrontierMath Tiers 1-3Accuracy51.2%Independentxhigh12 Jun 2026Epoch AI
FrontierMath Tiers 1-3Accuracy24.6%Independentlow28 Aug 2026Epoch AI
FrontierMath Tiers 1-3Accuracy17.2%Independentnone28 Aug 2026Epoch AI
FrontierMath Tiers 1-3 (v1)Accuracy28.3%Independenthigh15 Apr 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy88.9%Independentxhigh7 Aug 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy87.2%Independenthigh15 Apr 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy26.7%Independentnone7 Aug 2026Epoch AI

Coding & software engineering

Coding & software engineering results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
FrontierCodeMain score (Mean@5)27.0%Third partyunknown—https://cognition.com/frontiercode via Epoch AI
WeirdML (v2)Average accuracy60.3%Third partyhigh—https://htihle.github.io/weirdml.html via Epoch AI
WeirdML (v2)Average accuracy37.9%Third partynone—https://htihle.github.io/weirdml.html via Epoch AI

Agentic tool use

Agentic tool use results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Deep Research Bench (FutureSearch)Average score36.3%Third partylow—https://evals.futuresearch.ai/#drb via Epoch AI

Factuality & hallucination

Factuality & hallucination results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
SimpleQA VerifiedAccuracy (% correct)29.4%Independenthigh27 Aug 2026Epoch AI

Human preference

Human preference results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Code Arena: WebDev (arena.ai)Arena score (Elo scale)1397Independenthigh25 Sep 2026Arena
Code Arena: WebDev (arena.ai)Votes12008Independenthigh25 Sep 2026Arena
Text Arena (arena.ai)Arena score (Elo scale)1447Independenthigh25 Sep 2026Arena
Text Arena (arena.ai)Votes63140Independenthigh25 Sep 2026Arena
Vision Arena (arena.ai)Arena score (Elo scale)1251Independenthigh28 Sep 2026Arena
Vision Arena (arena.ai)Votes25775Independenthigh28 Sep 2026Arena

Other GPT-5 versions

Sources