Kaiju-Coder is a model I trained myself with MLX on Apple Silicon. No reseller, no wrapper. It runs on a Mac Studio in my office and answers an OpenAI-compatible endpoint you can point your code at right now.
$ base_url = "https://api.rmdw.ai/v1" · drop-in, streaming, zero migration
No OpenAI key behind a paywall. No fine-tune of a hosted base you don't control. Kaiju-Coder was trained with MLX on Apple Silicon, quantized to run fast on consumer hardware, and is served from a machine sitting in my office. That is the entire pitch, and it is true.
Built from the ground up with Apple's MLX framework on Apple Silicon. The weights are mine. The training data and recipe are tuned for one job: shipping small-business software.
No cloud GPU bill, no third-party inference vendor. A single Mac Studio (M3 Ultra, 256 GB) in Atlanta runs the model and answers the API. You can watch the host name in every response.
Same request shape, same response shape, streaming included. Change one base_url and your existing OpenAI code talks to Kaiju-Coder. No rewrite, no SDK swap.
Grab a key, point your client at the endpoint, and send a message. It speaks the OpenAI Chat Completions format, so anything that already works with OpenAI works here.
curl https://api.rmdw.ai/v1/chat/completions \
-H "Authorization: Bearer rmdw-YOUR_KEY" \
-H "Content-Type: application/json" \
-d '{"messages":[{"role":"user","content":"Build me a landing page for a bakery"}]}'# pip install openai. it's a true drop-in
from openai import OpenAI
client = OpenAI(
base_url="https://api.rmdw.ai/v1",
api_key="rmdw-YOUR_KEY",
)
resp = client.chat.completions.create(
model="kaiju-coder-mlx",
messages=[{"role": "user",
"content": "Build me a landing page for a bakery"}],
)
print(resp.choices[0].message.content)// npm i openai
import OpenAI from "openai";
const client = new OpenAI({
baseURL: "https://api.rmdw.ai/v1",
apiKey: "rmdw-YOUR_KEY",
});
const resp = await client.chat.completions.create({
model: "kaiju-coder-mlx",
messages: [{ role: "user", content: "Build me a landing page for a bakery" }],
});
console.log(resp.choices[0].message.content);# streaming works exactly like OpenAI. just pass stream=True
stream = client.chat.completions.create(
model="kaiju-coder-mlx",
messages=[{"role": "user", "content": "Write a Stripe invoice webhook"}],
stream=True,
)
for chunk in stream:
print(chunk.choices[0].delta.content or "", end="")Your welcome email gives you a key. Point any OpenAI-compatible client at the endpoint with these two values and you are live. Nothing new to install.
Use the OpenAI library you already have. Set base_url to the endpoint and api_key to your key. The quickstart above has copy-paste examples in curl, Python, and Node.
In the tool's model settings, set the OpenAI base URL to https://api.rmdw.ai/v1 and paste your key. Your AI coding editor now runs on Kaiju-Coder.
Add a custom OpenAI connection with the same base URL and key. You get a full chat interface on the model, nothing to build.
Kaiju-Coder isn't trying to win a general benchmark. It's specialized for the business-builder niche: the websites, tools, and plumbing that small companies actually pay for. That's the work it does at RMDW, every day.
Clean, responsive marketing sites and landing pages from a one-line brief. HTML/CSS/JS, React, or Next. Shipping-ready, not toy output.
Stripe checkout, invoice generators, webhook handlers, billing logic. Knows the patterns small businesses need to actually get paid.
Contact forms, intake flows, lead-gen widgets wired to email and a database. The front door of any small-business funnel.
Contact databases, deal stages, follow-up automations. The glue that turns a list of leads into tracked, closeable revenue.
Schedulers, dashboards, quote calculators, internal admin panels. The unglamorous custom software that runs an operation.
Serverless functions, integrations, scheduled jobs. The wiring between the tools a business already pays for.
Start free, upgrade when the work is real. Every plan hits the same model on the same hardware. You're just buying more of it.
Need volume beyond Pro, or a private model tuned for your business? That's RMDW consulting. Email info@rmdwllc.com.
One subscription gets you a key (prefixed rmdw-) for a model I built, on hardware I own. Same OpenAI-compatible endpoint, same quickstart above.
Checkout takes a minute. Your API key is emailed to you, usually within a few hours.