Covenant Labs Platform
Private OSS LLM Platform
Covenant Labs Platform
Covenant Labs Platform is the fastest way to deploy private open‑source LLMs for testing, experimenting, and production.
Quick Start
Launch a private model
Model
Llama 3.1 8B
Context
128k
Status
Ready
Price/hr
Calculated
Fast Iteration
Experiment quickly find which model works best for your problem.
Transparent pricing
We calculate the exact compute required for your model and let you choose the compute platform that makes the most sense.
Private by default
Deploy inside environments with audit‑friendly logs and no vendor data retention.
Transparent pricing
No black boxes. Enter your model specs — we’ll calculate VRAM and show GPUs that fit, with clear per-hour pricing.
- Exact compute breakdown
- Predictable per-hour pricing, terminate any time
GPU
NVIDIA L4 (24GB)
$/hr
$0.60
GPU
NVIDIA A100 (80GB)
$/hr
$2.30
GPU
NVIDIA H100 (80GB)
$/hr
$4.10
Estimates shown for demo purposes. Final pricing is computed from your model and chosen infrastructure.
Experimental end-to-end model encryption
Protect weights at rest and in use through our Model Encryption Protocol and the Covenant Labs Python SDK. Our experimental secure inference keeps your keys under your control while enabling OpenAI-compatible requests.
- ✓Hold your own secrets — we never see your keys.
- ✓Compatible with OpenAI-style APIs for seamless swap-in.
- ✓Deploy in stages: encrypt → deploy → ready.
How experimental secure inference works
- Encrypt model weights locally using keys you manage.
- Deploy to your chosen compute; keys never leave your boundary.
- Send requests via an OpenAI-compatible endpoint; responses are processed securely.
- Audit-friendly logs with no vendor data retention.
This flow lets teams experiment with private OSS LLMs while maintaining strict control over secrets and artifacts — ideal for pilots, sensitive evaluations, and staged rollouts.
Experimental feature — availability may vary by model and environment.