a year ago, everyone was debating GPT-4 vs claude 3.

that was the whole conversation.
two models. one fight. simple.
now i'm staring at a list of ten serious coding models from six different countries, and half of them are names most western founders still can't pronounce correctly.
kimi K2.5. qwen 3 coder. GLM-4.7.
these aren't experiments. these are production-grade models eating real workloads.
the moat everyone thought openAI and anthropic had?
it's not gone. but it's a lot narrower than it looked 18 months ago.
the real story isn't which model wins.
it's that the race just got ten times harder to predict.. and the builders who ignore that are going to be very surprised very soon.
which one are you actually shipping with?