Capability-first sizing

Liwen Guesstimator

Estimate which Liwen edition, architecture, and hardware profile fits a customer use case. Start from the capability the customer needs, then review the recommended platform and optional technical sizing.

1 · Choose the Liwen edition

Light / Medium / Full

Liwen Light

local / edge

Single workload close to the data, without depending on large GPU servers. Suitable for focused reasoning systems such as Zen Shield backup/storage troubleshooting or Sentinella Hub Kubernetes troubleshooting.

Recommended for one primary use case
Focused troubleshooting assistants such as Zen Shield or Sentinella Hub
Best for branch, lab, factory, air-gapped site
Single-node Kubernetes is acceptable

Liwen Medium

team / business unit

Focused AI platform for specific workloads, teams or departments.

Multiple related workloads
Several models and agents
Shared RAG and business workflows
Small Kubernetes cluster or GPU server pair

Liwen Full

multi-workload / MoE

Complete sovereign AI platform for multiple workloads and Mixture-of-Experts models.

Multiple teams and tenants
MoE and large model capable
High availability and governance
Dedicated Kubernetes, GPU and data layers

2 · Select Liwen pillars

scope

Models as a Service

Local inference, curated model catalog, routing and model APIs.

Agents

Agent runtime, MCP gateway, tool calling and workflow orchestration.

Data / RAG

Document ingestion, embeddings, vector DB and knowledge retrieval.

3 · Choose expected usage

simple customer inputs

Usage pattern

Named users

Document corpus

Availability target

4 · Select target hardware

Dell / NVIDIA / custom

Hardware target

Deployment preference

Tenancy

Technical sizingShow technical sizing

Reveal CPU, RAM, GPU, storage, and sizing assumptions.

5 · Select desired model catalog

Liwen curated models

Detailed rationale

why this recommendation fits

Recommended edition

Hardware fit

Workload fit

Upgrade path

next recommended step

Liwen readiness score

Recommended architecture

Supported model fit

Model	Category	Status	Comment

Supported capabilities

Capability	Estimate	Notes

Hardware recommendation

Layer	Recommended footprint	Notes

Technical details

for architects

Resource	Guesstimate	Comment