>>105655293
my preset has 25000 in 2000 max out (27000 total)
with deepseek that's $0.0168/req
if you add +500 output tokens for the reasoning block that becomes $0.017875/req
assuming ideal cache conditions and ignoring first expensive write request, claude is $0.0375/req
over 2x from that alone
slowburner suffering
i hope i did my math right