← Back to home
Product · Models
Run any model. Your hardware, your call.
Run open models entirely on your machine, import any GGUF from Hugging Face, and reach for a frontier cloud model only when you want depth. Nupick matches models to your hardware and routes each task by privacy and complexity.
Local models, on-device
Run open models like Qwen, Llama, Gemma, Phi, and Mistral entirely on your machine. Fully offline when you want it — no account, no upload.
Import any GGUF
Pull any GGUF model straight from Hugging Face. Nupick builds it into your local library and it's ready to chat.
Cloud models on tap
When a task needs more depth, route to a frontier cloud model — using only sanitized, redacted context, never raw files.
Per-task routing
Stay Local, let Nupick choose with Auto, or go Cloud — per task. Private work stays on-device by default.
Hardware-aware recommendations
Nupick detects your CPU, GPU, and RAM and recommends models that actually fit — with a VRAM budget you can see before you download.
Embeddings & retrieval, local
On-device embeddings and reranking power grounded answers — your knowledge is indexed and searched on your machine.
Quantization
Pick the trade-off that fits your machine.
When you import a model, Nupick offers the available quantizations and defaults to the best balance of quality and size for your hardware — so a model that's too big to run never lands on your disk.
The principle
Local when it can. Cloud when it helps. Always your choice.
You should never be locked into one model or one place to run it. Nupick keeps the full range — small on-device models to frontier cloud ones — under your control, and tells you exactly what stays local.
Early access
Start building your
private AI workspace.
Local-first, privacy-first, and built for people who want powerful AI without handing over everything.