Released date

Model

Last Updated 5/22/2025

anthropic/claude-sonnet-4-20250514-thinking

Claude Sonnet 4 (Thinking)

Anthropic's latest-generation workhorse model, offering a balance of performance and speed.

Released Date: 5/22/2025

Avg. Accuracy:

70.6%

Latency:

221.85s

Performance by Benchmark

Benchmarks

Accuracy

Rankings

FinanceAgent

44.5%

( 5 / 33 )

44.5%

5 / 33

CorpFin

67.3%

( 11 / 44 )

67.3%

11 / 44

CaseLaw

74.0%

( 8 / 27 )

74.0%

8 / 27

TaxEval

75.9%

( 16 / 61 )

75.9%

16 / 61

MortgageTax

62.5%

( 27 / 35 )

62.5%

27 / 35

Math500

93.8%

( 11 / 55 )

93.8%

11 / 55

AIME

76.3%

( 15 / 51 )

76.3%

15 / 51

MGSM

90.9%

( 24 / 54 )

90.9%

24 / 54

LegalBench

81.3%

( 16 / 76 )

81.3%

16 / 76

MedQA

92.7%

( 12 / 57 )

92.7%

12 / 57

GPQA

74.5%

( 11 / 53 )

74.5%

11 / 53

MMLU Pro

83.8%

( 8 / 51 )

83.8%

8 / 51

MMMU

74.9%

( 10 / 32 )

74.9%

10 / 32

LiveCodeBench

62.4%

( 21 / 53 )

62.4%

21 / 53

IOI

4.6%

( 7 / 12 )

4.6%

7 / 12

Academic Benchmarks

Proprietary Benchmarks (contact us to get access)

Cost Analysis

Input Cost

$3.00 / M Tokens

Output Cost

$15.00 / M Tokens

Input Cost (per char)

N/A

Output Cost (per char)

N/A

Performance by Benchmark

Cost Analysis

Join our mailing list to receive benchmark updates on