Astropedia on-device models
ExecuTorch (.pte) builds of Saga, the Vedic-astrology assistant of the
Astropedia app: Gemma 3 270M fine-tuned for English, Hindi and Bengali,
8-bit quantized (8da8w; trained with 4-bit QAT), 2048-token context. The app downloads these files once and
runs them on the phone.
manifest.json lists every version with sizes and SHA-256; the app verifies
each file before using it.
Model catalog
manifest.json also lists the models the app offers in Settings ->
On-device model -> Change model (catalog). Only one runs at a time.
Models other than Saga are downloaded from their publishers' repositories
(react-native-executorch's ExecuTorch exports), pinned by commit and
checked by SHA-256; they are not redistributed here.
| Model | Adapter | Download | Languages | Licence |
|---|---|---|---|---|
| Astropedia Saga v2.1 | gemma21 |
199 MB | en best, hi best, bn best | Gemma Terms of Use |
| Qwen3 0.6B | instruct |
517 MB | en good, hi basic, bn basic | Apache 2.0 |
| Llama 3.2 1B | instruct |
1146 MB | en good, hi fair, bn basic | Llama 3.2 Community License |
| Qwen3 1.7B | instruct |
1315 MB | en good, hi fair, bn fair | Apache 2.0 |
Built with Llama. Llama 3.2 is licensed under the Llama 3.2 Community License, Copyright © Meta Platforms, Inc. All Rights Reserved.
Gemma is provided under and subject to the Gemma Terms of Use found at https://ai.google.dev/gemma/terms
- Downloads last month
- 46