your own AI.
nobody else's.
the okay model is a private AI model you pay for and run as your own: in the cloud just for you, on your own servers, or on your own computer. it speaks the OpenAI API, so okaydesigner's AI team (and anything else) can use it with one key.
cloud
a private endpoint on okaydeploy that only your keys can call. its own GPU, not shared with anyone. scales to zero when you're not using it. no prompts or answers kept.
from $49 a month
on-prem
runs on your own Linux servers with NVIDIA GPUs. one line installs it with docker compose. the licence is checked offline: nothing leaves your network.
$1,500 per GPU a year
download
runs on your Mac with Apple silicon. works on a plane. okaydesigner finds it on your computer.
$120 a year
what it is
the okay model is Qwen3-Coder-30B-A3B-Instruct from Alibaba's Qwen team, released under the Apache 2.0 licence, served for you by okaydeploy with an okay preset for product design. it's strong at tool calling and code, it's fast (3 billion of its 30 billion parameters work on each word), and its licence lets us host it for you and lets you run it yourself.
try it
curl https://okaydeploy.com/v1/chat/completions \
-H "authorization: Bearer okm_…" -H "content-type: application/json" \
-d '{"model":"okay-model","messages":[{"role":"user","content":"design a pocket synth"}]}'
sign in, make a key, and paste it into okaydesigner: connect an AI → okay model. docs