Legal
Open-source notices and model licenses
The third-party software Noema is built with, and the terms each Noema model is released under.
Software
The following table lists the main third-party components used by Noema.
| Component | Purpose in Noema | License |
|---|---|---|
| llama.cpp (ggml-org/llama.cpp) | Provides the C/C++ inference engine for GGUF models and vision support | MIT license |
| MLX (ml-explore/mlx) | Apple-optimized array framework for running MLX models | MIT license |
| Liquid AI SLM / LFM models | Small-language models used as an optional backend | LFM Open License v1.0 - based on Apache 2.0 with a $10 million revenue threshold for free commercial use |
| Open Textbook Library | Provides openly licensed textbooks for retrieval-augmented answers | CC BY 4.0 (open-textbook content is licensed to be freely used and adapted) |
| Hugging Face Hub | Source of downloadable AI models | Model-specific licenses (varies); API usage is public |
| SearXNG | Provides web search for tool-calling | AGPL-3.0 license |
| ggml library | Tensor library included via ggml headers (part of llama.cpp) | MIT license |
Noema app source
A community source edition of the Noema Apple client is published at github.com/noemaai-labs/noema under the Apache License 2.0, as stated in that repository's LICENSE file. It is a deliberately limited edition, not the production App Store source tree, and the code license does not grant rights to the Noema name or logos.
Model licenses
Noema models are published on Hugging Face. They are open weights, not open source: the weights are publicly downloadable, but no open-source license is granted. The terms for each release are on its model card.
Base models
Every Noema model is post-trained from a base model released by the Qwen team: Qwen3.5-2B for Noema 1.5 2B and the deprecated Noema 2B, and Qwen3.5-4B for Noema 1.5 4B. Both base models are published under the Apache License 2.0 on Hugging Face. We thank the Qwen team for releasing them openly.
Additional credits
- Noema Early Testing Team: For their dedicated testing and valuable feedback.
- Mr. Enes Deniz: For idea suggestions and sound effects for UX regarding accessibility.