--- license: other license_name: sam-license license_link: LICENSE base_model: facebook/sam3 tags: - mirror - cuvis-ai - sam3 - segmentation --- # SAM 3 (Segment Anything with Concepts) checkpoint (mirror) Byte-identical mirror of [`facebook/sam3`](https://huggingface.co/facebook/sam3) at revision `3c879f39826c281e95690f02c7821c4de09afae7` for the [Cuvis.AI](https://github.com/cubert-hyperspectral/cuvis-ai) plugins. No fine-tuning, conversion or re-serialization: the files below are the upstream release as published, with the upstream licence text in `LICENSE`. ## Why this mirror exists Cuvis.AI provisions model weights once into a shared cache (`download-model download ` from `cuvis-ai-core`) and then runs its pipelines in an offline, token-free runtime that only reads that cache. Hosting the exact upstream files under the `cubert-gmbh` organisation makes that provisioning reproducible (commit-pinned and sha256-verified) and takes the per-user Hugging Face account and token out of the users' path. ## Files and provenance | File | Upstream source | Upstream revision / id | sha256 | Size | |---|---|---|---|---| | `sam3.pt` | `facebook/sam3` / `sam3.pt` | `3c879f39826c281e95690f02c7821c4de09afae7` | `9999e2341ceef5e136daa386eecb55cb414446a00ac2b55eb2dfd2f7c3cf8c9e` | 3.45 GB | | `config.json` | `facebook/sam3` / `config.json` | `3c879f39826c281e95690f02c7821c4de09afae7` | `4616385e4b21f2e5e22c875b65679185cbccfa95de42542b9166f7dc3d57160f` | 25.8 kB | | `LICENSE` | `facebook/sam3` / `LICENSE` | `3c879f39826c281e95690f02c7821c4de09afae7` | `bec48f70bd37bf8280a9d1ebf01642d26086f6122ba735baa08fe03c5a6e7448` | 7.4 kB | Mirrored on 2026-09-04 by Cubert GmbH from the sources above. The sha256 values are the upstream values; `tools/mirror_weights.py check` in `cuvis-ai-core` re-verifies this repository against them and against the upstream licence text. ## Licence These files are Meta's "SAM Materials" and are distributed under the **SAM License** (Last Updated: November 19, 2025), reproduced verbatim in `LICENSE`. By downloading or using them you agree to be bound by that Agreement. Section 1.b.i allows redistribution only under the same terms and with a copy of the Agreement, which is what this repository does. Obligations that travel with the files: - You may not use, or permit others to use, the SAM Materials for any activities subject to the International Traffic in Arms Regulations (ITAR) or end uses prohibited by Trade Controls, including those related to military or warfare purposes, nuclear industries or applications, espionage, or the development or use of guns or illegal weapons (section 1.b.v). You must not be the target of Trade Controls, and your use must comply with Trade Controls and applicable laws, including privacy and data-protection laws (sections 1.b.iii and 1.b.v). - Publications of research performed with the SAM Materials must acknowledge their use (section 1.b.ii). Reverse engineering the SAM Materials is not permitted (section 1.b.iv). - The SAM Materials are provided "as is" without warranties (section 3); Meta may modify the Agreement (section 8), so check the upstream text for the current version. - If you allege in litigation that the SAM Materials or their outputs infringe intellectual property or other rights you own or can license, your licence terminates on the day the claim is filed, and you indemnify Meta against third-party claims arising from your use or distribution of the SAM Materials (section 5.b). On termination you must delete and cease use of the SAM Materials (section 6). The upstream repository `facebook/sam3` is gated; this mirror is ungated. Cubert GmbH redistributes these files as a Licensee under section 1.b.i and is itself bound by the Agreement, including the indemnity in section 5.b, for its own distribution. You are bound directly and separately by the same Agreement, including sections 5.b and 6 above. Cubert GmbH is not affiliated with or endorsed by Meta. ## Usage with Cuvis.AI ```text uv run download-model download sam3 ``` `download-model` (from `cuvis-ai-core`) provisions the file into the shared Hugging Face cache and verifies its sha256. From cuvis-ai-core 0.16.0 on, `ModelWeights.resolve("")` returns the cached path and the `cuvis-ai-sam3` nodes resolve their weights through it, so a pipeline needs no manual step once the weights are provisioned; earlier plugin releases still fetch from their original upstream sources.