TypeSeer

Acknowledgments

What TypeSeer is built on

TypeSeer stands on open-source software and open-weight AI models, and we are grateful to their authors. This page is the same attribution notice that ships inside the app, under About TypeSeer → Acknowledgments.

The full license texts are published here: every license, in full. A copy of each also ships inside the app, as the licenses require.

The AI models

TypeSeer uses Gemma 4 as its default on-device model, downloaded on first use after the user confirms the choice. The underlying model is © Google; “Gemma” is a trademark of Google LLC, and Digital Llama is not affiliated with, endorsed by, or sponsored by Google.

  • Google publishes Gemma 4 itself under the Apache License 2.0.
  • The built-in models TypeSeer downloads are Animateus/gemma-4-E2B-GGUF, the default, and Animateus/gemma-4-E4B-GGUF, offered on Macs with more memory. Each is a 6-bit quantization of Google’s release, made with llama.cpp. An install still running the earlier MLX conversion, Animateus/gemma-4-E2B-6bit, keeps it until it moves on its own. TypeSeer passes on the Gemma Terms of Use and Google’s Gemma Prohibited Use Policy for these models, and our Terms require you to comply with that policy: Gemma Terms of Use · Prohibited Use Policy.
  • TypeSeer also lets you download and run other models from Hugging Face. Each is governed by its own license, published on that model’s page, and you are responsible for reviewing and complying with it. TypeSeer grants you no rights in third-party models.

Open-source software

Every package TypeSeer builds on, with its copyright holder and SPDX license identifier.

  • llama.cpp - © 2023–2026 The ggml authors - MIT
  • mlx-swift - © 2023 ml-explore - MIT
  • mlx-swift-lm - © 2024 ml-explore - MIT
  • swift-transformers - © 2022 Hugging Face SAS - Apache-2.0
  • swift-huggingface - © 2025 Hugging Face SAS - Apache-2.0
  • swift-jinja - © 2022 Hugging Face SAS - Apache-2.0
  • Sparkle - © 2006–2017 Andy Matuschak, Elgato Systems GmbH, Kornel Lesiński, Mayur Pawashe, C.W. Betts, Petroules Corporation, Big Nerd Ranch - MIT, plus the bundled bsdiff (BSD-2-Clause), sais-lite (MIT), Ed25519 (Zlib) and SUSignatureVerifier (BSD-2-Clause) notices reproduced in its license file
  • swift-argument-parser - © Apple Inc. and the Swift project authors - Apache-2.0 with Runtime Library Exception
  • swift-asn1 - © 2022 The SwiftASN1 Project - Apache-2.0
  • swift-atomics - © Apple Inc. and the Swift project authors - Apache-2.0 with Runtime Library Exception
  • swift-collections - © Apple Inc. and the Swift project authors - Apache-2.0 with Runtime Library Exception
  • swift-crypto - © 2019 The SwiftCrypto Project - Apache-2.0
  • swift-nio - © 2017–2018 The SwiftNIO Project - Apache-2.0
  • swift-numerics - © Apple Inc. and the Swift project authors - Apache-2.0 with Runtime Library Exception
  • swift-syntax - © Apple Inc. and the Swift project authors - Apache-2.0 with Runtime Library Exception
  • swift-system - © Apple Inc. and the Swift project authors - Apache-2.0 with Runtime Library Exception
  • EventSource - © 2025 Mattt - MIT
  • yyjson - © 2020 YaoYuan - MIT

TypeSeer also uses Apple’s own frameworks (AppKit, SwiftUI, Foundation, NaturalLanguage, Accessibility, Metal), which are covered by the Apple SDK license rather than by any of the above.

TypeSeer for Windows

TypeSeer for Windows uses the same Gemma 4 model and llama.cpp, listed above, with llama.cpp’s Vulkan backend running the model on your graphics card. It also includes:

The installer adds Microsoft .NET (MIT) if your PC doesn’t have it. TypeSeer for Windows also uses features of Windows itself (UI Automation, text recognition, the spell checker and screen capture), which are covered by your Windows license.

Spotted something missing or misattributed? Tell us and we’ll fix it.