From the archive
Nebius AI Studio adds vision models, embeddings, and adapter hosting
Nebius expands its hosted model catalog beyond text, with image understanding, embeddings, and LoRA support.
Nebius AI Studio announced an expanded model catalog on December 17, 2024. The update added vision-language models, more language models, embeddings, and support for hosting LoRA adapters.
Images alongside text
The vision additions included Qwen2-VL and LLaVA models for tasks such as image captioning and visual questions. The text catalog also grew with models from the Llama, Qwen, and Phi families.
Embeddings supplied another building block for applications that search or group text by meaning. LoRA hosting offered a way to use model adaptations without treating every deployment as a full separate model.
A shared hosted interface
The models were available through Nebius's unified API, with base and fast inference options. The playground provided a route for testing them before integrating an application. The update broadened the hosted platform beyond its original language-model offering.