supermarket

Vision-language

GPT-5.4

By OpenAI, in the GPT-5 family.

Released
5 Mar 2026
Status
Legacy
Context
1.1M
Max output
128K
Input per 1M tokens
$2.50
Output per 1M tokens
$15.00
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 & reasoning42.5313 of 4
Coding & software engineering32.983 of 5
Agentic tool use35.1Not ranked1 of 2
Vision & multimodal understanding10.7Not ranked1 of 3
Factuality & hallucination45.1321 of 1
Human preference50.9274 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 solved44.0%Independentxhigh11 Mar 2026Epoch AI
Chess PuzzlesPuzzles solved38.0%Independentmedium15 Jul 2026Epoch AI
Chess PuzzlesPuzzles solved38.0%Independenthigh15 Jul 2026Epoch AI
Chess PuzzlesPuzzles solved20.0%Independentlow15 Jul 2026Epoch AI
Chess PuzzlesPuzzles solved5.0%Independentnone15 Jul 2026Epoch AI
DTBench (Decision Theory Benchmark)Accuracy (capability questions)94.4%Third partyxhigh—https://conceptualreasoning.ai/dtbench via Epoch AI
Earthborne Rangers (EBR-bench)Objectives reached25.4%Independentxhigh25 Jun 2026Epoch AI
LMCA (Language Model Conceptual Argumentation)Correlation with expert ratings (Pearson r)0.5197Third partyxhigh—https://conceptualreasoning.ai/lmca via Epoch AI
Mystery Game PuzzlesPuzzles solved37.0%Independentxhigh24 Jul 2026Epoch AI
Mystery Game PuzzlesPuzzles solved28.0%Independentmedium28 Aug 2026Epoch AI
Mystery Game PuzzlesPuzzles solved17.0%Independentlow27 Aug 2026Epoch AI
Mystery Game PuzzlesPuzzles solved16.0%Independentnone30 Aug 2026Epoch AI

Scientific reasoning

Scientific reasoning results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
GPQA DiamondAccuracy93.3%Independentxhigh6 Mar 2026Epoch AI
GPQA DiamondAccuracy89.9%Independenthigh15 Jul 2026Epoch AI
GPQA DiamondAccuracy88.9%Independentmedium15 Jul 2026Epoch AI
GPQA DiamondAccuracy84.8%Independentlow15 Jul 2026Epoch AI
GPQA DiamondAccuracy74.7%Independentnone15 Jul 2026Epoch AI

Mathematics

Mathematics results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
FrontierMath Tier 4Accuracy49.0%Independentxhigh11 Jun 2026Epoch AI
FrontierMath Tier 4 (v1)Accuracy27.1%Independentxhigh6 Mar 2026Epoch AI
FrontierMath Tiers 1-3Accuracy78.6%Independentxhigh11 Jun 2026Epoch AI
FrontierMath Tiers 1-3 (v1)Accuracy47.6%Independentxhigh6 Mar 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy97.8%Independenthigh15 Jul 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy95.6%Independentmedium15 Jul 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy95.3%Independentxhigh6 Mar 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy84.4%Independentlow15 Jul 2026Epoch AI
OTIS Mock AIME 2024-2025Accuracy57.8%Independentnone15 Jul 2026Epoch AI

Coding & software engineering

Coding & software engineering results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
DeepSWEPass@151.8%Third partyxhigh—https://deepswe.datacurve.ai/ via Epoch AI
GSOOPT@131.4%Third partyxhigh—https://gso-bench.github.io/index.html via Epoch AI
GSOOPT@125.5%Third partyhigh—https://gso-bench.github.io/index.html via Epoch AI
METR Time HorizonsAverage task success (Epoch's column)74.3%Third partyxhigh—https://metr.org/blog/2025-03-19-measuring-ai-ability-to-complete-long-tasks/ via Epoch AI
METR Time HorizonsAverage task success (Epoch's column)74.3%Source not statedunknown—Epoch AI via Epoch AI
MirrorCodeScore15.6%Independenthigh10 Aug 2026Epoch AI
PostTrainBenchAverage score19.0%Source not statedhigh—Epoch AI via Epoch AI
SWE-bench Verified% resolved76.9%Independenthigh6 Mar 2026Epoch AI
Terminal-Bench 2.0Accuracy (mean over trials)81.8%Third partyunknown—https://www.tbench.ai/leaderboard/terminal-bench/2.0 via Epoch AI
WeirdML (v2)Average accuracy77.7%Third partyxhigh—https://htihle.github.io/weirdml.html via Epoch AI
WeirdML (v2)Average accuracy57.4%Third partynone—WeirdML Leaderboard via Epoch AI

Agentic tool use

Agentic tool use results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Deep Research Bench (FutureSearch)Average score35.1%Third partylow—https://evals.futuresearch.ai/#drb via Epoch AI

Professional knowledge work

Professional knowledge work results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
APEX-AgentsPass@1 (all rubric criteria met)52.4%Source not statedunknown—Epoch AI via Epoch AI

Long context

Long context results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
CL-benchTasks solved27.9%Source not statedxhigh—Epoch AI via Epoch AI
CL-bench LifeTasks solved21.7%Source not statedxhigh—Epoch AI via Epoch AI
CL-bench LifeTasks solved19.3%Source not statedhigh—Epoch AI via Epoch AI
CL-bench LifeTasks solved13.8%Source not statedunknown—Epoch AI via Epoch AI

Vision & multimodal understanding

Vision & multimodal understanding results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Furniture AssemblyScore37.5%Independentxhigh10 Sep 2026Epoch AI

Factuality & hallucination

Factuality & hallucination results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
SimpleQA VerifiedAccuracy (% correct)45.1%Independentxhigh27 Aug 2026Epoch AI

Human preference

Human preference results
BenchmarkMetricValueProvenanceSettingEvaluatedSource
Code Arena: WebDev (arena.ai)Arena score (Elo scale)1394Independent—25 Sep 2026Arena
Code Arena: WebDev (arena.ai)Votes3484Independent—25 Sep 2026Arena
Document Arena (arena.ai)Arena score (Elo scale)1470Independent—13 Sep 2026Arena
Document Arena (arena.ai)Votes33331Independent—13 Sep 2026Arena
Text Arena (arena.ai)Arena score (Elo scale)1475Independenthigh25 Sep 2026Arena
Text Arena (arena.ai)Arena score (Elo scale)1465Independent—25 Sep 2026Arena
Text Arena (arena.ai)Votes67632Independent—25 Sep 2026Arena
Text Arena (arena.ai)Votes64479Independenthigh25 Sep 2026Arena
Vision Arena (arena.ai)Arena score (Elo scale)1284Independenthigh28 Sep 2026Arena
Vision Arena (arena.ai)Arena score (Elo scale)1279Independent—28 Sep 2026Arena
Vision Arena (arena.ai)Votes28881Independenthigh28 Sep 2026Arena
Vision Arena (arena.ai)Votes22854Independent—28 Sep 2026Arena

Other GPT-5 versions

Sources