EdgeDox brings offline PDF Q&A to Android

EdgeDox matters less as a polished app today than as proof that useful document Q&A is starting to fit inside a phone. If the team can keep latency, memory use, and model loading under control, this kind of offline assistant could become one of the clearest wins for edge AI.

• –Running Qwen3.5-0.8B on Android with quantized weights shows mobile LLM inference is moving from novelty to usable product territory
• –Document Q&A is a strong first use case because PDFs are privacy-sensitive and users have a clear reason to avoid cloud uploads
• –MNN is a key part of the story here since efficient mobile inference frameworks matter as much as the model itself
• –The main limitation is still the classic small-model tradeoff: tighter memory budgets mean weaker long-context handling and lower answer quality on complex documents