Edge-1.3B
A 1.3 billion parameter language model trained entirely from scratch.
Built on consumer Apple Silicon. No cloud GPU rentals. No pre-trained weight inheritance. Every parameter learned from zero.
Overview
Edge-1.3B is the flagship model of the Edge series — a family of sovereign language models developed by AXe Technologies. Designed for edge deployment where privacy, latency, and data sovereignty are non-negotiable.
| Parameters | 1,344,374,784 |
| Architecture | Proprietary transformer, optimized for edge inference |
| Training | From scratch — no fine-tuning, no transfer learning |
| Hardware | Apple Silicon (Metal acceleration) |
| Precision | Mixed precision (float16/float32) |
| Status | Pre-training in progress |
Design Philosophy
Edge models are built on three principles:
- Sovereignty — Trained on our own hardware, on our own terms. No cloud dependency.
- Efficiency — Optimized for deployment on consumer-grade devices, not datacenter GPUs.
- Privacy — Runs entirely on-device. Your data never leaves your machine.
The Edge Series
| Model | Parameters | Status |
|---|---|---|
| Edge-Medium | 85M | ✅ Complete |
| Edge-1.3B | 1.3B | 🔄 Training |
| Edge-3 | Planned | Architecture phase |
Intended Use
- On-device inference for privacy-sensitive applications
- Edge computing environments with limited connectivity
- Sovereign AI deployments requiring full data control
- Research into efficient transformer architectures
Limitations
Edge-1.3B is currently in active pre-training. Performance benchmarks will be published upon training completion. This model is not yet suitable for production deployment.
Training
Trained from scratch using proprietary infrastructure on Apple Silicon. Training methodology, data composition, and architectural details are not publicly disclosed.
"Given both the competitive landscape and the strategic value of sovereign AI infrastructure, this model card contains no further details about the architecture or training pipeline."
About AXe Technologies
Canadian AI research lab focused on sovereign, privacy-first artificial intelligence. We build models that run on your hardware, trained on our hardware. No cloud. No compromise.
Open to collaboration — Contact us for evaluation access and partnership inquiries.
Built in Canada 🍁 on Apple Silicon