Skip to content
Earlyn

Licenses

Last updated September 28, 2026

Models

Earlyn can run two models with its built-in engine. Neither comes inside the app: each is downloaded through earlyn.app when you press Download in Settings, and runs only on your Mac. The same texts are in the app under Settings, General, About, Licenses.

  • EmbeddingGemma 300MGemma Terms of Use

    The “Search by meaning” model. Downloaded by the app only when you press Download.

    Gemma Terms of UseGemma Prohibited Use PolicySource

  • Whisper large-v3-turboMIT

    The “Meeting transcription” model that writes down meetings in any language. Downloaded by the app only when you press Download. Copyright OpenAI; ggml conversion by the whisper.cpp authors.

    MIT LicenseSource

  • Qwen3 1.7BApache License 2.0

    The “Earlyn model” that answers, summarizes and writes digests. Downloaded by the app only when you press Download. Copyright Alibaba Cloud; an unmodified GGUF conversion.

    Apache License 2.0Source

Gemma: what you agree to

EmbeddingGemma is provided by Google under the Gemma Terms of Use. We pass those terms on to you: by downloading the model you agree to them, including the Gemma Prohibited Use Policy. The app shows this and asks before the download starts.

  • You may not use the model to break the law, or to harm, harass, defraud or deceive people.
  • You may not use it to generate or spread malware, spam or other abusive content.
  • You may not use it to make automated decisions that affect people's rights without human review.
  • If you pass the model file on to someone else, you must pass these terms on with it.
  • The model comes without warranty. Google may update the terms; the current text is on Google's page.

This is a summary to help you read the terms, not a replacement for them. Earlyn itself uses the model for one thing: turning your own memories into vectors so you can search them by meaning, on your Mac.

Open-source software