Pricing
Pay for the GPUs, not a markup on them.
Your cloud provider bills GPU hours to your own account at its list or negotiated price. AWSBee charges a flat fee for the control plane, with no per-token margin.
Developer
For individuals evaluating self-hosted models.
$0forever
- Up to 2 active deployments
- Unlimited gateway requests
- Playground and analytics (7 days)
- vLLM and NVIDIA NIM engines
- Community support
Team
For teams running models in production.
$49per month
- Unlimited deployments
- 90-day analytics retention
- Multiple cloud accounts
- Email support, 1 business day
- Usage exports
Enterprise
For regulated workloads and large GPU fleets.
Custom
- Self-hosted gateway in your VPC
- SSO and audit logs
- Private networking options
- Architecture reviews
- Dedicated support channel
What does a deployment cost to run?
The GPU instance is the main cost. A Llama 3.1 8B deployment on a single NVIDIA L4 (g6.xlarge) is roughly $0.80 per hour on demand in us-east-1; a 70B model on 4× L40S (g6e.12xlarge) is roughly $10.50 per hour. Storage for model weights (200 GiB gp3 by default) is billed while the stack exists, even when the instance is stopped. Use the GPU calculator to compare instance types for your model.