← All comparisons

Claude 4.5 Haiku (Non-reasoning) vs Gemma 3 12B Instruct

Anthropic vs Google — side-by-side benchmark comparison

Claude 4.5 Haiku (Non-reasoning)Gemma 3 12B Instruct
Intelligence Index31.08.8
Coding Index29.66.3
Math Index39.018.3
Output speed (tok/s)105.80.0
Blended price ($/1M)$2.19$0.14
Time to first token (s)0.65s0.00s
aime22.0%
aime 2539.0%18.3%
artificial analysis coding index29.606.30
artificial analysis intelligence index31.008.80
artificial analysis math index39.0018.30
gpqa64.6%34.9%
hle4.3%4.8%
ifbench42.0%36.7%
lcr43.7%6.7%
livecodebench51.1%13.7%
math 50085.3%
mmlu pro80.0%59.5%
scicode34.4%17.4%
tau232.5%10.8%
terminalbench hard27.3%0.8%

Benchmark data from Artificial Analysis.