0

Claude Sonnet 4.5

Closed

anthropic/claude-sonnet-4.5 is an AI model.

Open
Publisher
Anthropic
Family
Claude
Params
Undisclosed (closed)
Context
200K
Input
Output
Providers
anthropicaws-bedrockvertex-ai
License
Proprietary
Released
29 Sep 2025
Canonical
anthropic.com/news/claude-sonnet-4-5

Cite

Notes

Only stored in your browser.

Attribution

Attribution policy →
Context
200K
Input
$3.75
/1M tok
Output
$15
/1M tok
Blended
$6.56
#391/ 414/1M · 3:1
Speed
49
#185/ 414tok/s

Price & speed via Artificial Analysis · ranked across tracked models

Reported on 80 evals across 14 domains - best 87.5% on IDE-Bench · #1 of 13

Reported eval scores

80
AIME2025100
Spiral Bench0
NanochatAquaRat0
Infraresolutionbench83.8
Pii Masking0
Mastermind0.7
Mastermind0.7
MulSeT0
Taxcalc Bench0
Terminal Bench 2 Verified50.3
Traveling Salesman0
Extract Zero0
Mbti0
Taxcalc Bench0
Lights Out0
Gridgame0
Gridsize0
Count Intersections0
Vpct 10
Count Lines0
Hello World Multi Turn0
Futurex Past0
Pydantic Adherence0
Acereason Math0
Ascii Tree0
Reverse Text0
Meow0
Ai2Arc0
LiveBench - Coding80.4
LiveBench - Language76.5
Terminal-Bench (Hard)28.8
LiveBench - Reasoning77.6
IFBench42.7
LiveBench - Data Analysis57
SWE-bench Multilingual67
LiveBench - Math79.3
LiveBench - Instruction Following53.4
IDE-Bench87.5
Vibe Code Bench v1.122.6
MMLU-Pro86
LiveBench70.7
Humanity's Last Exam (HLE)7.1
SciCode42.8
GPQA Diamond72.7
GPQA (Full Set)83.9
SWE-bench Verified74.8
SWE-bench77.2
LiveCodeBench59
MedScribe84.1
MedCode40.6
τ²-bench (Tau²-bench)70.5
OSWorld-Verified62.9
TaxEval v273.3
CorpFin v260.8
MortgageTax64
ProofBench19
AIME 2025: Problems from the American Invitational Mathematics Examination37
FrontierMath15.2
GSM8K0
MATH-5000
LegalBench0
SAGE32.9
PubMedQA: A Dataset for Biomedical Research Question Answering0