Superlinked's SIE serves the models behind agent tasks. LM-Kit One serves the tasks themselves: documents, search, and agents, with a production control plane.
Both are self-hosted servers for the models agents call. They answer at different altitudes of the same stack.
Superlinked SIE
Four model primitives, encode, score, extract, and generate, behind OpenAI-compatible endpoints, serving a 100+ model catalog on demand. Apache 2.0, Docker and Kubernetes native, with real depth in retrieval models: dense, sparse, and multi-vector.
LM-Kit One
Document intelligence, extraction, search with citations, and governed agents as ready endpoints, plus OpenAI, Anthropic, Ollama, MCP and native dialects on one engine, with identities, policies, audit, and an admin console built in.
The rows that decide real deployments, not a feature checklist.
| Dimension | LM-Kit One | Superlinked SIE |
|---|---|---|
| Primary job | Application backend: documents, search, and agents as endpoints, plus model serving | Model serving for agents: encode, score, extract, generate |
| API surface | OpenAI, Anthropic, Ollama, MCP, native REST | OpenAI-compatible endpoints |
| Document intelligence | Full pipelines: OCR, Markdown, splitting, extraction with confidence, redaction, signatures, PDF/A | Serves document-to-markdown models; the pipeline stays in your code |
| Search and grounded answers | Built-in service: ingestion, hybrid retrieval, reranking, answers citing document and page | Serves dense, sparse, and reranker models; the retrieval stack is yours to assemble |
| Agents | Server-side agents with skills, governed tools, memory, MCP | Serves the models an agent loop calls; the loop lives in your app |
| Governance and operations | Admin console, identities, SSO, per-key grants, audit, capability policies | Not documented; typically fronted by your own gateway |
| Platforms | Windows, Linux, macOS; installers, desktop mode, Windows service | Linux and macOS, via Docker, Kubernetes, or native Python |
| Hardware | CPU-first; CUDA, Vulkan, and Metal acceleration | CPU or NVIDIA CUDA; MLX on Apple Silicon |
| Scaling out | Horizontal scaling, any node serves any request, KEDA-ready | Kubernetes-native: Helm, KEDA scale-from-zero, Terraform modules |
| Licensing | Free to build and evaluate; Professional for larger production use | Apache 2.0 open source; managed cloud offered separately |
SIE is young and moving fast; this table reflects our reading of its public documentation at publication. Check their site for the current state. Corrections are welcome through contact.
One question settles most cases: do you want model primitives to build on, or the built application services to call?
Choose Superlinked SIE
You are building the retrieval or agent stack yourself, want raw model primitives on your own GPUs and Kubernetes, and prefer an Apache 2.0 engine you can read and fork.
Choose LM-Kit One
You want document intelligence, cited answers, and governed agents as endpoints applications call today, on Windows, Linux, or macOS, with identities, audit, and operations built in.
Teams that want the same engine inside their own process embed LM-Kit.NET: One and the SDK are one engine in two delivery forms.
LM-Kit One
