Our scrubbing & audit layer, running Sigmix-authored firmware — with an optional PCIe card for the appliance.
Explore Sigmix Guard
Every request served on-box, sealed from the internet. KSA Cloud runs elastic and in-Kingdom.
See your postureA sovereign route served only inside the Kingdom, and an international route for frontier models. Each key belongs to one of them, and a sovereign request never crosses over.
New accounts get 7.50 ($ 2.00) of free credit, valid for 30 days. No card.
// OpenAI-compatible · the gateway runs in Riyadh
# Sovereign — served only in the Kingdom https://inference.sigmix.sa/ksa-sovereign-route-only/v1 # International — frontier models, served abroad https://inference.sigmix.sa/international-route-only/v1
Both speak the same API, bill from the same credit and live in the same console. What differs is where a request may be answered.
For data that must stay in the Kingdom: government, health, finance, anything regulated.
/ksa-sovereign-route-only/v1For the frontier models that cannot run inside the Kingdom, when the data is allowed to travel.
/international-route-only/v1A residency claim is only as strong as the weakest place it could fail. So it is enforced in three independent places.
Each key is minted for one route. A sovereign key presented on the international path is refused before anything else happens.
Each route has its own base path and its own handler. Nothing in a request can choose the other one.
A sovereign model cannot be linked to an endpoint outside the Kingdom. The database refuses the row, whoever tries.
When every in-Kingdom endpoint for a model is down, the answer is an error — not a quiet detour through another country.
Five OpenAI-compatible endpoints on each route. The table is the live roster — models appear here when they are published.
| Endpoint | KSA only | Served abroad |
|---|---|---|
Chat | 7 models | 7 models |
Embeddings | — | — |
Text to speech | 2 models | — |
Speech to text | 2 models | — |
Image generation | — | — |
A dash means no model is published there yet, not that the endpoint is missing.
Before a request reaches a model, the most it could cost is held from your credit. A request you cannot pay for is refused, not run.
When it finishes you are charged what it used, from the model's own count, and the rest of the hold comes back.
Metered in US dollars and shown in either at the fixed 3.75 peg. When you top up, the tax invoice is in riyals with VAT.
If your code already calls OpenAI, change the base URL and the key. That is the whole migration.
curl https://inference.sigmix.sa/ksa-sovereign-route-only/v1/chat/completions \
-H "Authorization: Bearer $SIGMIX_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"command-a-03-2025","messages":[{"role":"user","content":"Summarise Vision 2030 in one sentence."}]}'import os
from openai import OpenAI
client = OpenAI(base_url="https://inference.sigmix.sa/ksa-sovereign-route-only/v1", api_key=os.environ["SIGMIX_API_KEY"])
reply = client.chat.completions.create(
model="command-a-03-2025",
messages=[{"role": "user", "content": "Summarise Vision 2030 in one sentence."}],
)
print(reply.choices[0].message.content)import OpenAI from 'openai';
const client = new OpenAI({ baseURL: 'https://inference.sigmix.sa/ksa-sovereign-route-only/v1', apiKey: process.env.SIGMIX_API_KEY });
const reply = await client.chat.completions.create({
model: 'command-a-03-2025',
messages: [{ role: 'user', content: "Summarise Vision 2030 in one sentence." }],
});
console.log(reply.choices[0].message.content);Make a key for each route and call them. The console shows every request as it lands.