← All comparisons

Claude 4.5 Sonnet (Non-reasoning) vs DeepSeek V3.2 Speciale

Anthropic vs DeepSeek — side-by-side benchmark comparison

Claude 4.5 Sonnet (Non-reasoning)DeepSeek V3.2 Speciale
Intelligence Index37.129.4
Coding Index33.537.9
Math Index37.096.7
Output speed (tok/s)54.70.0
Blended price ($/1M)$6.56$0.00
Time to first token (s)1.08s0.00s
aime
aime 2537.0%96.7%
artificial analysis coding index33.5037.90
artificial analysis intelligence index37.1029.40
artificial analysis math index37.0096.70
gpqa72.7%87.1%
hle7.1%26.1%
ifbench42.7%63.9%
lcr51.3%59.3%
livecodebench59.0%89.6%
math 500
mmlu pro86.0%86.3%
scicode42.8%44.0%
tau270.5%0.0%
terminalbench hard28.8%34.8%

Benchmark data from Artificial Analysis.