API
< Back to all clusters
[TECHNOLOGY] · France · 6 sources

started · updated

OVHcloud unveils Olala, its first European AI model family

French cloud provider OVHcloud has announced Olala, its first family of large language models (LLMs) designed for business use. The debut model, Olala mini 7A1B, is an open-weight model featuring a 7-billion parameter architecture with a mixture-of-experts (MoE) design. This allows only 1 billion parameters to be active per query, optimizing memory usage and speed.

Developed by the OVHai LLM laboratory, the model was trained using approximately 7 million euros worth of computing time on European supercomputers, including Jupiter, MeluXina, and Jean Zay. The model is designed to be highly efficient, capable of running on standard CPUs without a dedicated graphics card or on GPUs with as little as 8 GB of memory.

OVHcloud is also launching OVHai Workspace, an agentic AI platform for businesses that includes productivity, web AI, and knowledge management tools. To strengthen the European AI ecosystem, OVHcloud is collaborating with partners such as Qwant, Ecosia (via EUSP), and Shadow. The search engine Qwant has already integrated the Olala mini model into its Flash Answer feature, reportedly improving response speeds by 20 percent.

Entities

Europe · Hugging Face · JUPITER supercomputer · MeluXina supercomputer · OVHcloud · Octave Klaba · Olala · Olala mini · Qwant