Sageling
Free local AI assistant for Apple Silicon Macs that transcribes, researches and drafts entirely offline
🔗 Visit Sageling
Description
Most AI assistants send everything you type to a company's servers. Sageling flips that: it downloads a language model onto your own Mac and runs it there, so a lawyer's case notes or a doctor's dictation never leave the laptop.
Sageling is a free, local-only AI assistant for Apple Silicon Macs (M1 and later) built on Qwen 3.5, a 9-billion-parameter open-weight model that runs fully offline after the initial ~8GB download. It handles transcription from audio and video, document and spreadsheet comparison, web research, drafting, and file organization, aimed at professionals in healthcare, legal, education and other fields where sending data to a cloud AI is a non-starter. Because the model is small enough to run on a laptop GPU, usage is unlimited with no rate limits or subscription — but it also means noticeably weaker reasoning than GPT-4 or Claude, more factual errors, and real battery/fan-noise cost while it's working. It requires 16GB+ RAM.
💬 Our review
The short version: Sageling is a solid pick if the whole point is that your data never leaves your machine, and a poor pick if you actually need frontier-model reasoning quality.
Compared to ChatGPT or Claude, you're trading a meaningfully smaller and less capable model for complete privacy and zero usage limits, on hardware you already own. Compared to Ollama (bring-your-own local model), Sageling packages a specific model with task-oriented features (transcription, doc comparison, drafting) instead of a raw runtime, which is friendlier if you don't want to pick and tune a model yourself. It's free, so there's no pricing risk — the real cost is accepting a 9B model's mistakes and the fan noise of running inference locally. Skip it if your work benefits from a state-of-the-art model and you're fine with cloud AI; pick it if privacy is non-negotiable and Mac + Apple Silicon is your platform.
📊 Global score
🤖 AI-enriched data
100% gratuit, aucun palier payant. Modèle open-weight (Qwen 3.5), aucun coût de licence.
Pros
Confidentialité totale — aucun envoi de données vers un cloud par défaut
Usage illimité, aucune limite de requêtes ni abonnement
Fonctionne hors ligne après l'installation initiale
Gère transcription, comparaison de documents, recherche web et rédaction
Cons
Réservé aux Mac Apple Silicon (M1 et plus) — pas de Windows, Linux ou Intel Mac
Modèle 9B nettement moins fiable que GPT-4 ou Claude sur du raisonnement complexe
Consommation GPU intensive : bruit de ventilateur et batterie sollicités, 16 Go de RAM minimum