Models & quality

NELA offers Fast, Smart, Deep, and Auto so you can trade speed for depth without hunting model names every time.

On this device (Private)

Install models in Settings, then map Fast / Smart / Deep to the ones you like.

Typical families:

• Chat models — conversation and writing.
• Vision — image questions.
• Speech — listen and talk (TTS / ASR).
• Retrieval helpers — improve document and File Indexer quality (optional advanced installs).

Start with one solid chat model. Add vision or speech only when you need those modes.

NELA Cloud

When Cloud is on, tiers use hosted models. Switching Fast / Smart / Deep does not download a local model file. Smart and Deep usually spend credits; see Pricing and Cloud settings.

Auto

On Cloud, Auto can pick a tier for you. On this device, Auto typically uses your Smart local model.

Tips

• Smaller models are snappier on laptops; larger ones need more RAM.
• Compatibility hints in Settings help you avoid downloads that will not fit.
• Artifacts and long decks often feel better on Cloud Smart/Deep, but Private still works.

Next: Private vs Cloud · Get started