Enterprise
Decisions that stay in your estate.
Run our models on the servers, laptops and GPUs you operate. On your own hardware, decisions are made there and nothing is sent to us.
- 1,400held-out decisions
- 91.8%correct
- 73%kept at a 5% target
- 2.7%wrong among those kept
- 27%to a person
Run it on hardware you own.
Your servers
Our models run on the servers you already operate, including disconnected networks.
Laptops and phones
The small models need only a CPU, so they run on the devices your people already use.
One GPU you own
Myles-G runs on a single GPU you own. It is available by request.
The answers a security review needs.
Data stays put
When our models run on your hardware, decisions are made there and nothing is sent to us.
Works offline
The models need no network connection, so they run on air-gapped networks.
Published lineage
Each new model ships with a record of its base model, data and licences.
Measured errors
A model keeps a decision only when it clears a certified error rate on held-out data. The rest go to a person.
UK public text
New models are trained on legislation, GOV.UK guidance and the parliamentary record, under open licences.
Built in the UK
Trained and evaluated on UK-owned hardware.
Pay for a licence, not for traffic.
Evaluate on your decisions
Send a sample of the decisions you want to automate. We measure our models on them and show the error rate and what would go to a person.
One licence per deployment
Monthly or annual, with no per-decision fees. Tell us what you need and we will send a quote.
Upgrade with the evidence
New versions come with their measured results, so you can compare them on your own decisions before you switch.
Talk to us.
Pilots, self-hosting and security questions. Tell us what you need and we will reply by email.