The Model Index
Up to four models

GPT-5.4 vs GPT-5.2

On the index (GPT-4 = 100), GPT-5.4 scores 166 and GPT-5.2 162. Side by side on the same third-party tests, with price and context. The link in your address bar shares this exact comparison.

Next to GPT-5.4
Presets
GPT-5.4
166150–182
WeightsClosed
ReleasedMar 2026
Price in / out$2.5 / $15
Context1.05M
Core tests taken7 of 9
GPT-5.2
162146–178
WeightsClosed
ReleasedDec 2025
Price in / out$1.75 / $14
Context400K
Core tests taken6 of 9

GPT-5.4 and GPT-5.2 are too close to call (4 points apart, within the uncertainty), ahead on 5 of 6 shared core tests, at a similar price.

Test by testthird-party results

GPT-5.4GPT-5.2
0%25%50%75%100%REASONING & KNOWLEDGEGPQA Diamond (Epoch AI)GPQA Diamond (Epoch AI)Humanity's Last ExamHumanity's Last ExamGPQA Diamond (Vals AI)GPQA Diamond (Vals AI)MMLU ProMMLU ProMMMU ProMMMU ProCL-benchCL-benchEBR-benchEBR-benchSimpleQA VerifiedSimpleQA VerifiedMATHSFrontierMath (tiers 1-3)FrontierMath (tiers 1-3)AIME-style maths (OTIS mock)AIME-style maths (OTIS mock)Chess puzzlesChess puzzlesFrontierMath (2025 set)FrontierMath (2025 set)FrontierMath tier 4 (2025 set)FrontierMath tier 4 (2025 set)FrontierMath (tier 4)FrontierMath (tier 4)CODINGSWE-bench VerifiedSWE-bench VerifiedSWE-bench Pro (public set)SWE-bench Pro (public set)LiveCodeBenchLiveCodeBenchSWE-benchSWE-benchVibe Code Bench v1.1Vibe Code Bench v1.1GSO-BenchGSO-BenchMirrorCodeMirrorCodeWeirdMLWeirdMLAGENTSDTBenchDTBenchDeepResearch BenchDeepResearch BenchFurniture assemblyFurniture assemblyLMCALMCAMystery game puzzlesMystery game puzzlesPROFESSIONAL WORKCorpFin v2CorpFin v2LegalBenchLegalBenchMedCodeMedCodeMedScribeMedScribeMortgageTaxMortgageTaxTaxEval v2TaxEval v2NOVEL PROBLEMSARC-AGI-2ARC-AGI-2ARC-AGI-1ARC-AGI-1
Every result as a table
TestGPT-5.4GPT-5.2
Reasoning & knowledge
GPQA Diamond (Epoch AI)93.3%91.4%
Humanity's Last Exam36.2%27.8%
GPQA Diamond (Vals AI)91.7%91.7%
MMLU Pro87.5%86.2%
MMMU Pro87.5%86.7%
CL-bench27.9%18.2%
EBR-bench25.4%23.0%
SimpleQA Verified45.1%37.1%
Maths
FrontierMath (tiers 1-3)78.6%67.4%
AIME-style maths (OTIS mock)97.8%96.1%
Chess puzzles48.0%48.0%
FrontierMath (2025 set)47.6%40.7%
FrontierMath tier 4 (2025 set)27.1%18.8%
FrontierMath (tier 4)49.0%31.7%
Coding
SWE-bench Verified76.9%73.8%
SWE-bench Pro (public set)59.1%29.9%
LiveCodeBench84.1%85.4%
SWE-bench78.2%75.8%
Vibe Code Bench v1.167.4%53.5%
GSO-Bench31.4%27.5%
MirrorCode23.3%32.2%
WeirdML77.7%72.2%
Agents
DTBench94.4%90.9%
DeepResearch Bench35.1%41.1%
Furniture assembly37.5%38.3%
LMCA52.0%43.9%
Mystery game puzzles37.0%23.0%
Professional work
CorpFin v265.3%65.9%
LegalBench86.0%82.8%
MedCode41.3%49.8%
MedScribe77.5%84.4%
MortgageTax68.3%67.1%
TaxEval v274.0%75.8%
Novel problems
ARC-AGI-274.0%72.9%
ARC-AGI-193.7%94.5%
Task length · METR45 sec45 sec

Tests at least two of these models have taken. Bold rows are the core tests.

This page as a chartDownload PNG