Runs on this device Prompts and output stay in your browser.

Continue a passage

The most capable model here. Type a beginning and let it continue locally.

26/2000

Ctrl + Enter to load

Browser Lab

Settings

Real local inference via Transformers.js. Weights stay on your device.

Browser Lab

Runtime

Model
-
Backend
-
Precision
-
Prompt tokens
-
Generated
-
Load
-
First token
-
Prefill
-
Decode
-

Metrics reflect this device’s CPU/GPU runtime.

Browser Lab

Saved downloads

Models you load are saved in this browser’s Cache Storage. Unload frees memory; deleting here removes the files so they download again next time.

Checking saved files…

Browser Lab

Choose a model

Changing model unloads the current runtime, but your session history stays visible.

Selected model

Er-Medium

Largest lab model here. Still a tiny base completion model - unfactual and not a chatbot.

Parameters
12,497,520
Context
384 tokens
Download
~50 MB
Network
Weights only
Designed to stay local

AskJASP does not receive your prompt or continuation for generation. Hugging Face and the runtime CDN still receive normal download requests.

Know the model

Tiny base completion models can be false, incoherent, or inappropriate. They are not reliable assistants.

View on Hugging Face