The Model Index
Up to four models

o3 vs GPT-5.4 Mini

On the index (GPT-4 = 100), o3 scores 147 and GPT-5.4 Mini 147. 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 o3
Presets
o3
147131–163
WeightsClosed
ReleasedApr 2025
Price in / out$2.0 / $8.0
Context200K
Core tests taken6 of 9
GPT-5.4 Mini
147131–163
WeightsClosed
ReleasedMar 2026
Price in / out$0.75 / $4.5
Context400K
Core tests taken5 of 9

o3 and GPT-5.4 Mini are too close to call (0 points apart, within the uncertainty), ahead on 0 of 3 shared core tests, at 2.1× the price.

Test by testthird-party results

o3GPT-5.4 Mini
0%25%50%75%100%REASONING & KNOWLEDGEGPQA Diamond (Epoch AI)GPQA Diamond (Epoch AI)GPQA Diamond (Vals AI)GPQA Diamond (Vals AI)MMLU ProMMLU ProMMMU ProMMMU ProSimpleQA 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)CODINGLiveCodeBenchLiveCodeBenchWeirdMLWeirdMLAGENTSDTBenchDTBenchDeepResearch BenchDeepResearch BenchLMCALMCAMystery game puzzlesMystery game puzzlesPROFESSIONAL WORKCorpFin v2CorpFin v2MortgageTaxMortgageTaxTaxEval v2TaxEval v2NOVEL PROBLEMSARC-AGI-2ARC-AGI-2ARC-AGI-1ARC-AGI-1
Every result as a table
Testo3GPT-5.4 Mini
Reasoning & knowledge
GPQA Diamond (Epoch AI)81.8%86.9%
GPQA Diamond (Vals AI)84.1%83.1%
MMLU Pro85.6%84.5%
MMMU Pro80.4%79.3%
SimpleQA Verified49.4%29.4%
Maths
FrontierMath (tiers 1-3)33.3%51.2%
AIME-style maths (OTIS mock)84.4%88.9%
Chess puzzles37.0%23.0%
FrontierMath (2025 set)18.7%28.3%
FrontierMath tier 4 (2025 set)2.1%2.1%
Coding
LiveCodeBench83.9%81.5%
WeirdML52.4%60.3%
Agents
DTBench84.8%80.0%
DeepResearch Bench45.2%36.3%
LMCA39.7%40.8%
Mystery game puzzles29.0%11.0%
Professional work
CorpFin v259.7%60.9%
MortgageTax65.7%63.5%
TaxEval v274.6%71.2%
Novel problems
ARC-AGI-26.5%18.9%
ARC-AGI-175.7%63.7%
Task length · METR39 sec

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

This page as a chartDownload PNG