Question 1

What AI models does Axion AEOS run?

Accepted Answer

Axion AEOS is model-agnostic. It runs open-weight models locally (including Llama 3, Mistral, Qwen, DeepSeek, and specialized fine-tunes), as well as local embeddings and vision models. We select and quantize the model package based on your hardware budget and latency requirements.

Question 2

What hardware infrastructure do we need?

Accepted Answer

Hardware requirements depend on model size and concurrency. Typical deployments range from workstation-class single GPU setups (NVIDIA RTX 4090 / L40S) for small teams, up to multi-node H100 / A100 GPU clusters for enterprise data processing. We evaluate your hardware posture during the Strategy Session.

Question 3

Can Axion AEOS work fully air-gapped?

Accepted Answer

Yes. Axion AEOS is engineered for complete network isolation. It does not require outbound internet access for model inference, license telemetry, or operational tasks. System updates and model revisions are delivered via encrypted physical media packages.

Question 4

How do system updates and model revisions work?

Accepted Answer

For connected environments, updates are delivered via encrypted container releases that apply during scheduled maintenance windows. For air-gapped environments, updates are verified and applied offline without breaking system continuity.

Question 5

What happens if AvantVue disappears? Is there source code escrow?

Accepted Answer

Enterprise buyers frequently ask this about specialized software vendors. We offer source code and deployment container escrow options for enterprise contracts. Because you own the physical deployment inside your infrastructure, the system will continue running indefinitely on your hardware even without external support.

Question 6

Do we need in-house Machine Learning staff to operate it?

Accepted Answer

No. Axion AEOS is designed as a private operating system with automated model management, local memory decay, and built-in RBAC interfaces. Your standard IT or DevOps team can manage daily administration without specialized ML engineers.

Question 7

How does licensing work?

Accepted Answer

Deployments are licensed under a perpetual or annual private node model with fixed upfront pricing. We do not bill per API token, per document processed, or per user seat.

Technical & Enterprise FAQ

Have a specific infrastructure question?