Caleta Private AI Appliance
Caleta IT Solutions Limited · Operations & Productivity
Certification per Microsoft Marketplace.
Evidence tier Source Confirmed · 9 captures on record
What the publisher says
As described on Microsoft Marketplace.
Private AI does not need to cost thousands a month. The Caleta Private AI Appliance runs open-weight language models on Azure spot capacity, and tells you where to run them: the cheapest regions for your VM size, ranked by measured eviction rates. Spot economics without the roulette.
One Azure VM becomes a private model server inside your own tenant. SSH in, type llm, pick a model from a measured catalogue, and serve it through a bundled chat UI or an OpenAI-compatible API. Nothing leaves your subscription.
Show the rest of the publisher’s description (17 more lines)
Key Solution Features:
- Spot placement built in: a live ranking of the cheapest regions for your VM size, scored against real eviction rates over 24-hour, 7-day and 30-day windows. Azure does not publish this data; we measure it. Pick a region on evidence instead of guessing.
- Serving in about five minutes: deploy to first answer, model download included. We timed it. There is no stack to assemble.
- Downloads at hyperscale speed: the download path is tuned for the fastest possible pull from Hugging Face: multi-gigabit, measured at 30 Gbit/s on the largest sizes. Models that take an hour to fetch elsewhere land in minutes.
- Park it overnight: deallocate when idle and the only standing charge is the OS disk. On restart, or after a spot reclaim, the box re-downloads its model and serves again unattended.
- Measured, not promised: every model and VM size pairing on the menu was benchmarked on this exact image, with the exact serve settings shipped. Nothing is listed that has not been run. The current models and numbers are at https://caleta.io/appliance/guide/
- The menu stays current: new measured models publish to our catalogue service. Pull them onto a box you already deployed, no rebuild, no new image. It updates only when you ask it to.
- Two architectures: x86 sizes detect the GPU and serve through CUDA on NVIDIA or ROCm on AMD automatically. Azure Cobalt Arm sizes serve on CPU with one large memory pool. Frontier mixture-of-experts models run with no GPU at all, at single-user speeds. Yes, really.
- Private by default: chat UI and API bind to localhost, the host firewall is on, and a fresh API key is minted at every boot. Publish on :443 with an automatic certificate only when you choose to, with optional source IP restriction.
- Expert-friendly: every serve setting is visible and editable at serve time. The menu is the easy path, not a cage.
- No bundled weights: models download under their own licences, direct from Hugging Face. The image ships engines and tooling only.
Who it is for:
Enthusiasts, developers and teams who want open-weight models with full data sovereignty, without building the serving stack themselves.
Prerequisites:
You need an Azure subscription with vCPU or GPU quota for your chosen VM size. Setup help is in the online guide.
Getting started:
Deploy, SSH in, type llm. The trial model downloads in about a minute and the first chat is minutes away.
Preview
5 images
Agent build and provenance
See the full provenance
The layer-by-layer build, the evidence behind each claim, the risk basis and the cross-marketplace links are open to any account. Some rows are disclosed, some the source leaves Unknown; a free account shows you which.
Compliance
- FedRAMPConfirmedNot listed90%, registry-checkedNo FedRAMP Marketplace entry matched this vendor's domain, checked 2026-08-27registry recordas observed 2026-08-27
Confirmed means matched to a public authoritative registry. Claimed means the vendor or its listing states it, not yet cross-checked. A framework not shown was not found in any source we hold, which is not evidence against it. Not listed means a scoped registry check found no match for this vendor's domain: a No is a scoped registry check, not a compliance judgment. Confidence bands: 95% domain-verified, 90% registry-checked, 80% self-attested, 70% weak signal. Self-attested items marked “vendor's site” are gathered from the vendor's own website and are not verified by us.
Vendor
External enrichment
Sources
Publisher resources
3 linksLinked repositories
Unknown means this listing does not publish a repository. It is not a statement that the code is closed, and a linked repository is not a claim that the publisher wrote it: the registry computes that relationship privately and does not publish it.
Evidence risk is the share of the build you cannot see before you deploy, not a security rating. Sign in to see the layer-by-layer basis for this band.