LLMs

The LLMs layer is the model plane: engines that serve inference at scale and toolchains that fine-tune or adapt base models.

It is deliberately thin in this catalog. Most “agent” products sit in Agentic; this layer exists so serving (vLLM, TGI, llama.cpp, …) and fine-tuning (Unsloth, …) stay separately reviewable from orchestration frameworks.

Choose Serving for runtime inference; Fine-tuning when the cataloged pack is about adapting weights.