This page requires JavaScript to display.
Unpacking...
\n\n\n\n\n\n\n\n\n\n\n\n
\n\n \n \n\n \n
\n
\n
\n
the secure canadian AI
\n

Point your app
at the scAI

\n

An OpenAI-compatible endpoint on sovereign Canadian GPUs. Swap one base URL and your stack turns Canadian, more secure, and cheaper. Nothing else changes.

\n
Sovereign AI · Made in Canada
\n
\n Create account →\n See the swap\n
\n
\n
\n
\n\n \n
\n
\n
The only diff
\n

Keep your code. Move the endpoint.

\n
\n\n \n
\n
\n
Azure OpenAI
\n
\n
\n
\n
Gemini
\n
\n
\n
\n
AWS Bedrock
\n
\n
\n
\n\n
\n \n
\n
Before · {{ beforeLabel }}
\n 
{{ beforeCode }}
\n
\n \n
→
\n \n
\n
After · the scAI
\n 
from openai import OpenAI client = OpenAI(    base_url = \"https://api.thescai.ca/v1\",    api_key  = SCAI_API_KEY,)
\n
\n
\n

OpenAI-compatible. Same SDK, same calls. Point the client at api.thescai.ca.

\n
\n\n \n
\n
\n
Four numbers. No GPUs, no ops.
\n
\n \n
\n
Cost
\n
$3–6 /M
\n
output tokens. 60–80% below frontier APIs
\n
\n \n
\n
Throughput
\n
~2,000 tok/s
\n
sustained per GPU. Lower latency, in-country
\n
\n \n
\n
Security
\n
\n
\n
the scAI96
\n
\n
\n
\n
GPT-4 class94
\n
\n
\n
\n
Llama Guard81
\n
\n
\n
\n
adversarial block, WildGuard. Published
\n
\n \n
\n
Sovereignty
\n
100% CA
\n
weights + data in Canada. 0 foreign kill-switches
\n
\n
\n
\n
\n\n \n
\n
\n
Three steps
\n

Live in minutes.

\n
\n
\n
\n
01
\n
\n
Create an account
\n

No sales call. No cluster to provision.

\n
\n
\n
02
\n
\n
Get your key + endpoint
\n

An API key and an OpenAI-compatible URL.

\n
\n
\n
03
\n
\n
Swap the base URL
\n

Redeploy. You're sovereign.

\n
\n
\n
\n\n \n
\n
\n
\n
The portfolio
\n

One API. A neutral model line-up.

\n

Route by capability, cost, or sovereignty. One safety harness governs them all.

\n
\n
\n
\n
\n Flagship\n Intelligence 51\n
\n
GLM-5.2
\n

Top-scoring open model. Frontier-class reasoning, open-weight economics.

\n
\n
\n
\n Regulated default\n Intelligence 38\n
\n
Nemotron 3 Ultra
\n

NVIDIA, US-origin. The clean fit for sensitive public-sector buyers.

\n
\n
\n
\n Workhorse\n Intelligence 29\n
\n
Gemma 4 31B
\n

Google, open weights. Efficient, multilingual FR/EN. Carries the volume.

\n
\n
\n

Intelligence = score on the Artificial Analysis Intelligence Index for open-weight models. Higher is more capable; frontier closed models sit near 55.

\n
\n
\n\n \n
\n
\n
\n
\n
\n
\n
Sovereign by construction
\n

No foreign kill-switch.

\n

Open weights run on sovereign Canadian GPUs. Your data never crosses the border, and your endpoint answers to Canadian law alone. An API no one abroad can switch off.

\n

Every model is red-teamed and guardrailed, then scored against frontier models. You get the number, and can defend it to risk and legal.

\n
\n
\n\n \n
\n
\n
\n
Early access
\n

Bring your stack home.

\n

One base URL change. Sovereign by tonight.

\n
\n\n
\n
\n\n
\n
\n \n \n
\n
\n \n \n
\n
\n\n
\n \n \n
\n\n
\n \n \n
\n\n
\n \n \n
\n\n \n\n

No credit card required. Canadian data handling from day one.

\n
\n
\n
\n
\n\n \n
\n
\n
\n the scAI\n \n Secure Canadian AI\n
\n
© 2026 · Sovereign Model-as-a-Service
\n
\n
\n\n
\n\n\n\n\n\n\n"