Best Home AI Server: glossary
Last updated
Eight terms that decide which machine runs which model.
Olares One → Home AI server → Ollama · Open WebUI · Immich → Family AI that stays at home
VRAM
Memory on the graphics card. A model runs fastest when all of its weights fit in VRAM; whatever doesn't fit spills to sy
Unified memory
Memory shared by the CPU and GPU in one pool, as on Apple's M-series chips, NVIDIA GB10 (DGX Spark) and AMD Ryzen AI Max
Quantization
Storing a model's weights at lower precision, for example 4-bit instead of 16-bit, so it needs about a quarter of the me
Tokens per second
How fast a model writes its answer. A token is roughly three quarters of an English word; reading speed is around 5 to 1
Context window
How much text a model can consider at once: the prompt, documents and the conversation so far. Longer contexts need more
Mixture of experts (MoE)
A model built from many smaller expert networks, of which only a few run for each token. Qwen3-30B-A3B has about 30B par
CUDA
NVIDIA's platform for running software on its GPUs. Most AI tools, including vLLM and most image and video generators, a
Retrieval-augmented generation (RAG)
Letting a model answer from your own documents: the system searches the documents, passes the relevant passages to the m
Sources for this entry
- Olares One specifications (Olares docs) (Company-published)
- Olares One FAQ (Olares docs) (Company-published)
Continue this reading path
How the related websites connect
Cite this entry
Best Home AI Server. “Best Home AI Server: glossary.” Content reviewed 13 November 2026.
https://besthomeaiserver.com/glossary/