The Model Index
Up to four models

Claude Haiku 4.5 vs Claude 3.5 Haiku

On the index (GPT-4 = 100), Claude Haiku 4.5 scores 137 and Claude 3.5 Haiku 109. 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 Claude Haiku 4.5
Presets
Claude Haiku 4.5
137121–153
WeightsClosed
ReleasedOct 2025
Price in / out$1.0 / $5.0
Context200K
Core tests taken3 of 9
Claude 3.5 Haiku
10993–125
WeightsClosed
ReleasedNov 2024
Price in / out–
Context–
Core tests taken1 of 9

Claude Haiku 4.5 leads Claude 3.5 Haiku by 28 points, ahead on 1 of 1 shared core tests.

Test by testthird-party results

Claude Haiku 4.5Claude 3.5 Haiku
0%25%50%75%100%REASONING & KNOWLEDGEGPQA Diamond (Epoch AI)GPQA Diamond (Epoch AI)GPQA Diamond (Vals AI)GPQA Diamond (Vals AI)MMLU ProMMLU ProMATHSAIME-style maths (OTIS mock)AIME-style maths (OTIS mock)MATH (level 5)MATH (level 5)MGSMMGSMFrontierMath (2025 set)FrontierMath (2025 set)CODINGLiveCodeBenchLiveCodeBenchWeirdMLWeirdMLAGENTSBALROGBALROGDTBenchDTBenchPROFESSIONAL WORKCorpFin v2CorpFin v2LegalBenchLegalBenchTaxEval v2TaxEval v2
Every result as a table
TestClaude Haiku 4.5Claude 3.5 Haiku
Reasoning & knowledge
GPQA Diamond (Epoch AI)71.2%38.1%
GPQA Diamond (Vals AI)72.2%37.9%
MMLU Pro78.7%64.1%
Maths
AIME-style maths (OTIS mock)66.7%4.3%
MATH (level 5)96.4%46.4%
MGSM92.2%84.6%
FrontierMath (2025 set)5.9%0.3%
Coding
LiveCodeBench41.2%41.9%
WeirdML45.4%30.7%
Agents
BALROG31.2%19.3%
DTBench73.6%56.7%
Professional work
CorpFin v260.6%50.8%
LegalBench81.2%70.3%
TaxEval v267.5%57.4%

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

This page as a chartDownload PNG