- Services
MetaEncoder-30B is a multimodal embedding model from Meta AI, posted on Hugging Face on 2026-11-12. It maps text, images and video into 6,656-dimension L2-normalized vectors for retrieval, and the same vectors answer yes/no, multiple-choice and scored questions about a scene. It is built by contrastively fine-tuning the Muse-Glimmer-30B decoder, ships in BF16 Safetensors and fits on a single 80 GB GPU with transformers 5.15 or later. The model card carries no license text.
A situational 30B-parameter dense embedding model from Meta. Treat the modality benchmarks above as the leading indicator of fit — composite scoring across modalities is still maturing. Newly released, so production-readiness is still being shaken out.
Generated from this model’s benchmarks and ranking signals. Editor reviews refine it over time.
Access model weights, configuration files, and documentation.
No benchmark data available for this model yet.
Benchmark data sourced from: Hugging Face, Arena.ai, Artificial Analysis.
The top devices for this model at 4-bit, ranked by fit and speed.
| Device | Grade | VRAM |
|---|---|---|
| Acer Veriton GN100 AI MiniAcer | SS | 18.9 GB |
| AI TOP ATOM ATAGB10-9003Gigabyte | SS | 18.9 GB |
| AMD Instinct MI300XAMD | SS | 18.9 GB |
| AMD Instinct MI325XAMD | SS | 18.9 GB |
| AMD Instinct MI350PAMD | SS | 18.9 GB |
Cheapest current cloud rentals with at least 19 GB VRAM, refreshed hourly.
Advertising disclosure: we earn commissions when you shop through the links below.
| Option | Cost / GPU-hour |
|---|---|
NVIDIA GeForce RTX 3090Vast.ai · Spot · 24 GB VRAM | $0.11 |
NVIDIA RTX A5000RunPod · Community · 24 GB VRAM | $0.16 |
NVIDIA RTX PRO 4000 BlackwellVast.ai · Spot · 24 GB VRAM | $0.16 |
NVIDIA GeForce RTX 3090Vast.ai · On-Demand · 24 GB VRAM | $0.18 |
NVIDIA GeForce RTX 4090Vast.ai · Spot · 24 GB VRAM | $0.20 |
Per-GPU rate across RunPod, the Vast.ai marketplace, DigitalOcean, and Vultr.
Spot tier is interruptible. Plan for restarts when comparing against on-demand prices.
MetaEncoder-30B is a bi-encoder with a natural language interface: a task and each candidate are written in free-form text, with images and video as side information, and matching is an inner product over cached representations. The MetaEncoder wrapper exposes encode, retrieve, decide and calibrate calls. Decisions return a value, probabilities and a confidence score, with a default softmax temperature of 0.02. Images keep native resolution snapped to a 28-pixel grid up to 4,096 merged tokens, and video clips longer than 8 frames are sampled down to 8 frames. Training sequences were capped at 16,384 tokens, and nothing is truncated at inference.
Explore the Provider
Aggregate stats, leaderboard, release timeline, and benchmark coverage across every Meta model we track.
Check which of your devices can run this model and how fast.
Compare hosted per-token prices before you commit to local hardware.
Break-even math for self-hosting, renting a GPU, or paying per token.
Live status for the hosted APIs you might use instead of running this locally.
View on Hugging Face