From the model card
Fields below are copied from the tags and counters on the HuggingFace repository meta-models/Muse-Glimmer-30B at our last fetch. They are set by the uploader, not verified by us; rows with no tag are omitted. How this page is made.
- Publisher (HF namespace)
- meta-models
- Pipeline tag
- image-text-to-text
- Library
- Transformers
- Weight formats
- safetensors
- License tag
apache-2.0— read the license file in the repo before relying on it- Papers cited
- arXiv:2504.13181, arXiv:2602.06036
- Downloads (HF counter at last fetch)
- 623,035
- Likes (HF counter at last fetch)
- 1,845
- Model card
- https://huggingface.co/meta-models/Muse-Glimmer-30B
Use cases
- Multimodal chatbots processing natural images alongside user text
- Document understanding where visual layout and text are both present
- Research into 30B-scale multimodal architecture from the cited papers
- Self-hosted vision assistant to replace closed-source API endpoints
- Azure deployment for managed multimodal serving
Pros
- Apache 2.0 enables commercial self-hosting
- Methodology described in two arXiv papers (2504.13181, 2602.06036)
- Azure deploy integration simplifies managed cloud serving
- 30B scale provides meaningful step up from 7-13B VLM baselines
Cons
- Community org unaffiliated with Meta AI; users may misattribute model provenance
- No independent third-party evaluation vs competitive 30B VLMs published
- Requires roughly 60GB VRAM at BF16; GPU-heavy for on-premise deployment
- Safety alignment properties are not documented in the model card
Tags
transformerssafetensorsmuse_glimmerimage-text-to-textconversationalarxiv:2504.13181arxiv:2602.06036license:apache-2.0eval-resultsendpoints_compatibleregion:usdeploy:sagemaker