Tally โ€” on-device assistant model (Gemma 4 E2B, UD-Q4_K_XL)

This repo hosts the exact GGUF the Tally iOS app downloads (once, user-initiated) for its fully on-device receipt assistant.

Property Value
File Gemma4-E2B/gemma-4-E2B-it-UD-Q4_K_XL.gguf
Size 3,753,905,504 bytes
SHA-256 e19785539eb181af829dfb9f139111789a62e219d11cff52322d4c41d9ceb4e8
Base model Google Gemma 4 E2B (instruction-tuned), Unsloth-style UD-Q4_K_XL dynamic 4-bit quant
License Apache 2.0

The app pins an immutable commit revision of this repo and verifies the download byte-for-byte (size + GGUF magic + SHA-256) before use โ€” the file in this repo is intentionally never replaced in place. A model upgrade ships as a new file + a new pinned revision inside the app.

The download request carries no user data; receipts and questions never leave the device.

Downloads last month
7
GGUF
Model size
4B params
Architecture
gemma3n
Hardware compatibility
Log In to add your hardware

4-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support