Model Tracker
DINOv2
DINOv2 is an AI model made by Meta. Meta released it on 17 April 2023. It is generally available. Its weights are open. Its licence is Apache 2.0.
In inglese
- Released
- 17 April 2023 — ai.meta.com, read 16 October 2026
- Status
- Generally available — ai.meta.com, read 16 October 2026
- Takes
- Image — github.com, read 16 October 2026
- Returns
- Not confirmed: The model card says the model returns class, patch, and optionally register tokens. These feature tokens are not among the register's output categories (text, image, audio, video, code).
- Open weights
- Yes — github.com, read 16 October 2026
- Licence
- Apache 2.0 — ai.meta.com, read 16 October 2026
- Context window
- Not confirmed: The maker's model card and documentation reviewed do not state a context-window size for DINOv2.
- Longest answer
- Not confirmed: The maker's model card and documentation reviewed do not state a maximum answer or output-token limit for DINOv2.
- Knowledge cutoff
- Not confirmed: The maker's model card and documentation reviewed do not state a knowledge-cutoff date for DINOv2.
- API price, input
- Not confirmed: No standard API input-token price for DINOv2 is stated on the maker's pages reviewed.
- API price, output
- Not confirmed: No standard API output-token price for DINOv2 is stated on the maker's pages reviewed.
- Where
- Not confirmed: two readings disagree: app, weights and weights
- Parameters
- Not confirmed: DINOv2 is presented as a family of models with different parameter counts, not one count for the family: the maker lists ViT-S, ViT-B, ViT-L, and ViT-g variants.
- Model card
- https://github.com/facebookresearch/dinov2/blob/main/MODEL_CARD.md — github.com, read 16 October 2026
Questions
When was DINOv2 released?
Meta released DINOv2 on 17 April 2023.
Is DINOv2 still available?
Yes: it is generally available, as Meta states it.
Are the weights of DINOv2 open?
Yes: Meta publishes the weights of DINOv2 for download.
What is the licence of DINOv2?
The licence of DINOv2 is Apache 2.0, as Meta names it.
More from Meta
- Audiobox
- AudioGen
- Chameleon
- CM3leon
- Code Llama
- DINOv3
- Emu
- Emu Edit
- Emu Video
- EnCodec
- I-JEPA
- ImageBind
- LLaMA
- Llama 2
- Llama 3
- Llama 3.1
- Llama 3.2
- Llama 3.3
- Llama 4 Maverick
- Llama 4 Scout
- M2M-100
- Make-A-Video
- Movie Gen
- Multi-Band Diffusion
Sources
- DINOv2: State-of-the-art computer vision models with self-supervised learning (read 16 October 2026)
- Evaluating the fairness of computer vision models (read 16 October 2026)
- Model Card for DINOv2-S/B/L/g (read 16 October 2026)
- README.md (read 16 October 2026)