GPU Service
Two ways to get compute from hardware Lycoria owns and operates — rent a full machine outright, or work against our hosted models by the token. No shared tenants, no watered-down benchmarks.
Full GPU access
Rent a dedicated machine outright — your own environment, your own stack, nothing shared with other tenants. Billed by the hour, cancel any time.
- On-DemandPrice on request
- 1 MonthPrice on request
- 3 MonthPrice on request
How it works
- Tell us the workload and duration you're planning for
- We provision a dedicated B300 machine under your account
- You get full root/admin access — install anything, configure anything
- Billing runs by the hour for as long as the machine is yours
API access by the token.
Lycoria is a Malaysian-owned and operated company, running six models on our own infrastructure — performing the way each benchmark promised, not watered down by a shared public endpoint. No machine for you to provision or maintain.
- Six models, live
- Inference & tuning
- Token metering
- Usage & logs
How it works
- 1Get an API key — no machine to provision or maintain
- 2Call any of the six live models through a single endpoint
- 3Usage is metered per token, across whichever models you use
- 4We handle tuning, monitoring, and uptime on our side
The roster
Six models, tuned and monitored on our own infrastructure — each one performing the way its benchmark promised, not the way a shared, oversubscribed endpoint delivers it.
Not sure which one fits?
If you need full control over the environment — custom frameworks, specific driver versions, long training runs — Full GPU access gives you the whole machine. If you just need model output and don't want to manage infrastructure, Token as a Service gets you there faster. Tell us what you're building and we'll point you to the right one.