Alif Islamic — v4 Base (MediaPipe .task)
On-device LLM for the Alif Islamic app. This is a Qwen2.5-1.5B-Instruct
model packaged as a MediaPipe / LiteRT .task bundle for fully offline inference on mobile devices via
flutter_gemma.
Files
| File | Size | Description |
|---|---|---|
alif-islamic-v4-base.task |
~903 MB | MediaPipe LLM Inference task bundle (weights + tokenizer + metadata) |
Intended use
Answering Islamic questions (Quran, Hadith, fiqh) inside the Alif app. The model is scoped to only respond to Islamic topics and to refuse unrelated questions. It runs on-device for the offline path; an online path uses a larger hosted model for higher-quality answers.
⚠️ This model can be wrong on fiqh rulings. Answers are not a substitute for a qualified scholar.
Usage with flutter_gemma
import 'package:flutter_gemma/flutter_gemma.dart';
final modelManager = FlutterGemmaPlugin.instance.modelManager;
// Download the .task bundle (public repo — no token required)
await modelManager.downloadModelFromNetwork(
'https://huggingface.co/ahmedtamseer3/alif-islamic-v4-base/resolve/main/alif-islamic-v4-base.task',
);
final model = await FlutterGemmaPlugin.instance.createModel(
modelType: ModelType.gemmaIt, // generic IT chat template
maxTokens: 1024,
);
final chat = await model.createChat();
await chat.addQueryChunk(Message.text(text: 'What is the meaning of Surah Al-Fatihah?', isUser: true));
final response = await chat.generateChatResponse();
Direct download
https://huggingface.co/ahmedtamseer3/alif-islamic-v4-base/resolve/main/alif-islamic-v4-base.task
Base model & license
- Base model:
Qwen/Qwen2.5-1.5B-Instruct - License: Apache-2.0 (inherited from the base model)
- Format: MediaPipe LLM Inference
.task(LiteRT)
Limitations
- Small (1.5B) on-device model — limited reasoning vs. larger hosted models.
- May produce inaccurate fiqh rulings; use scholar-reviewed sources for authoritative answers.
- English/Arabic focused. Other Indian-language support is handled by a separate translation pipeline, not this model.