Enterprise
One way to measure AI compute, across every team
NEURAX gives research, engineering, platform and finance the same figures for the same model, computed before anything runs on machines they control—with the optional network paths made explicit.
Built for the organisation that pays the compute bill
01
Local processing, explicit network paths
The compiler, supported dataset profiling and training run on the selected machine. Importing a public model requests its configuration from Hugging Face or ModelScope; using the optional Agent sends its prompt and project design/data profile to the provider you configure. For strict boundaries, use a trusted internal endpoint or disable those optional connections.
02
Your model, your endpoint
Neurax AI talks to the provider you choose, or to a model hosted inside your own network. No key of ours, and nobody in between.
03
Figures a finance team can sign
Memory, time, energy and cost for each design, each with its basis (exact, measured or estimated), and what is unknown said as unknown, never drawn as zero.
04
Decisions on the record
The Decision workspace is the page someone signs: what a run commits, what could go wrong, what would change it, recorded against the figures it was taken on.
05
Regulation, tracked
Time Machine projects cost and carbon over the years ahead and tracks the regulation that applies to the model you built, from EU AI Act thresholds to carbon reporting.
06
Nothing locked in
Every run writes ordinary PyTorch files and weights into your own project folders. Stop using NEURAX and everything you trained keeps working.
From a first model to every team
- 01ScopeWe look at the models you train and the hardware they run on.
- 02PilotA team runs NEURAX on its own designs and checks the predictions against its own runs.
- 03Roll outVolume licences, onboarding for each team, and help with your security review.
Enterprise plan
Custom
For organisations standardising how AI compute is measured.
Everything in Pro, plus
- Volume licensing and invoicing
- The assistant on your internal model endpoint
- Onboarding for your teams
- Help with your security review
- A direct line to the people building NEURAX
Book a demo
See NEURAX on your own models
Tell us what you train and where it runs. We will walk you through NEURAX on an architecture like yours: memory, time and cost on the hardware you actually have.