You control your data. We never see it.
Beta Launching soon!
We're working hard to bring you our first beta. Sign up to be the first to know when we launch.
Expected in October 2025.
A secure, cloud-based AI chat platform designed for privacy, simplicity, and peace of mind.
Alpha testing has begun for our AI platform, with core features implemented and being refined for secure, private browser-based interactions—all managed on Canadian servers. This development log for version 0.7 highlights our commitment to privacy, security, simplicity, and working toward SOC 2 compliance.
Miigwech GPT (Mai) has reached the alpha stage. Core features are in place and undergoing active testing to provide private and secure AI interactions from your web browser, with all operations handled exclusively on secure servers located in Canada.
Enable Ghost Mode to prevent any data from being stored on your device and block analytical data from being collected.
Choose from various AI models by size and use case. *Only models up to 7B are available in alpha.
Our architecture uses distinct servers for training and production, eliminating the possibility of data leaks.
Experience responsive chat with AI models that generate answers in moments. The brief processing time ensures replies are accurate, secure, and timely.
All processing happens on secure servers located in Canada, ensuring data sovereignty.
Powered by NVIDIA A40 GPUs delivers up to 37.4 TFLOPS (FP32) and up to 149.7–299.4 TFLOPS. Estimated speeds:
Model | Tokens/s | 100T | 150T | 200T |
---|---|---|---|---|
3B | 60–70 | 1.4–1.7s | 2.1–2.5s | 2.9–3.3s |
7B | 45–50 | 2.0–2.2s | 3.0–3.3s | 4.0–4.4s |
14B | 44–45 | 2.2–2.3s | 3.4–3.5s | 4.5–4.6s |
32B | 23–26 | 3.8–4.3s | 5.8–6.5s | 7.7–8.7s |
* Figures are estimated
Our workflow uses LoRA for parameter-efficient fine-tuning and PyTorch for full model training.
3B - llama, gemma, qwen
7B - mistral, deepseek-coder, qwen2.5-coder
EE2E encryption, HTTPOnly cookies, and secure servers.
All servers and data processing are based in Canada for added privacy and compliance.
You own your data. All chat sessions are processed in real-time and not stored on our servers. Ghost Mode is available to prevent all but essential cookies.
Our models are routinely updated to deliver the latest capabilities, with a strict policy ensuring that no user data is ever retained between updates.