Nvidia woos wary Europeans with more open weights models
New embedding, parsing, and reranking models promise smarter agentic RAG workflows
Nvidia is trying to attract European customers who are cautious about AI by releasing new open weights models covering various sectors such as chatbots, healthcare, and robotics. The company has unveiled fresh Nemotron Retriever embedding and Alpamayo autonomous vehicle models at GTC Berlin. Embedding models are crucial for enterprise applications, particularly in retrieval augmented generation (RAG) technology.
Nvidia's embedding models support both image and text in 34 languages and are available in sizes ranging from one to eight billion parameters, with NVFP4-quantized versions for use on new Blackwell-based accelerators. For processing PDF documents, Nvidia has launched a refreshed Nemotron Parse model, which can handle about 10.5 pages per second on a single H100.
The new embedding models have been ranked among the top dense models on ViDoRe V3.1 benchmarks at BF16-precision. Additionally, Nvidia is offering access to Alpamayo 2 Edge, a more compact version of its autonomous driving model, designed for real-time processing of streaming video and sensor data on the DRIVE AGX Thor platform.
These models are unique as Nvidia releases both the weights and the training data used to create them, allowing customers to inspect the models and even build custom versions based on Nvidia's Nemotron architecture.
Written by urgent.news from The Register's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
Also reported by 1 other outlet
- Nvidia woos wary Europeans with more open weights models theregister.com
