LLMs

The LLMs layer is the model plane: Serving for inference at scale and Fine-tuning for adapting base weights. It is deliberately thinner than Agentic. Most “agent” products live above; this layer keeps serving and training reviewable on their own.

Choose Serving for tokens/sec and hardware utilization; Fine-tuning when domain language must live in weights. Gateways and agents call these endpoints — they should not replace them. Features such as Bring Your Own Model and Cost, Latency & Operational Economics are the layer’s falsifiers.

Representative packs vLLM, Ollama, and Unsloth span cluster serve, local serve, and efficient adaptation. Keep residency and license explicit; model choice is an ops decision with financing consequences when latency and spend slip.

Inline members for audit: vLLM Ollama Unsloth. Features: Bring Your Own Model Cost, Latency & Operational Economics. Related terms referenced: Agentic, Infrastructure.

For this research catalog the acceptance test is boring and strict: after wiring the pack, can a second person replay the same scenario and get the same measurable outcome? If not, the page is still a brochure. Document the scenario, the expected number, and the log line that proves it. That habit is what separates a deepened taxonomy cell from a landscape screenshot.

Also record what you refused: tools not connected, corpora not ingested, metrics not collected. Refusals keep the cell honest when marketing pages claim universality. A taxonomy essay that only lists winners teaches less than one that names the skip conditions next to the install path.

Keep the essay technical enough that an engineer can argue install order — and short enough that a reader finishes it before opening the right-rail member list. Prefer primary docs and repo evidence over landscape blurbs when the right-rail packs disagree.