← All comparisons

Phi-4 Mini Instruct vs Claude 4.5 Sonnet (Reasoning)

Microsoft vs Anthropic — side-by-side benchmark comparison

Phi-4 Mini InstructClaude 4.5 Sonnet (Reasoning)
Intelligence Index8.443.0
Coding Index3.638.6
Math Index6.788.0
Output speed (tok/s)0.055.0
Blended price ($/1M)$0.00$6.56
Time to first token (s)0.00s7.02s
aime3.0%
aime 256.7%88.0%
artificial analysis coding index3.6038.60
artificial analysis intelligence index8.4043.00
artificial analysis math index6.7088.00
gpqa33.1%83.4%
hle4.2%17.3%
ifbench21.1%57.3%
lcr13.7%65.7%
livecodebench12.6%71.4%
math 50069.6%
mmlu pro46.5%87.5%
scicode10.8%44.7%
tau28.2%78.1%
terminalbench hard0.0%35.6%

Benchmark data from Artificial Analysis.