OVHcloud launches European-trained open source AI model

The new lightweight LLM can run on standard desktop CPU, with plans to expand context window to 128,000 tokens

OVHcloud logo and branding pictured on a company data center in Roubaix, northern France.
(Image credit: Getty Images)

OVHcloud has launched a new family of open source AI models which have been exclusively trained on European compute infrastructure.

The new family of LLMs, dubbed Olala, includes the lightweight Olala mini 7A1B model, the first to be released. Olala mini 7A1B was trained exclusively on European supercomputers, including Jupiter, MeluXina, and Jean Zay.

According to OVHcloud, the model has been developed partly in anticipation of large tech companies pulling back from open source distribution models.

“AI is strategic for OVH,” said CEO Octave Klaba. “As well as offering AI services to our customers, we must be able to create models for our own internal needs and for those of our customers.”

Latest Videos FromIT Pro

“We cannot simply rely on the generosity of certain labs to share their LLMs as open source in the years to come,” Klaba added. “Know-how in model training has become widely accessible, and the cost of training has been divided by 20 in three years. It is now possible to create LLMs while making this activity profitable.”

Under the hood of OVHcloud’s new model

Built on a novel architecture, the model is designed to offer fast performance with lower memory requirements. According to the company, Olala mini 7A1B is able to run on a standard CPU at 50 tokens per second, and can run on an Nvidia RTX A4500 GPU at 160 tokens per second.

The model boasts 7 billion parameters, with a current context window of 32,000 tokens, although OVH expects the next version of the model to expand this window to 128,000 before the end of the year.

Olala mini 7A1B will be distributed on an open source basis with an Apache 2.0 license via Hugging Face, and will include model weights, technical specs and AI Act compliance documentation.

Olala mini 7A1B is already in use by French search engine Qwant, where it has reportedly delivered 20% faster responses than the model it replaced, with running costs three times lower.

OVHcloud plans to continue AI development with specialised models catering to enterprise-focused agentic AI workloads running on local devices, cloud DevOps workloads, and cybersecurity.

Sovereignty in the crosshairs

The announcement comes amid growing scrutiny on the relationship between European clients and US tech providers, with many EU-based organizations and governments seeking to decrease their reliance on American technology.

According to Gartner, while full digital sovereignty may not be a realistic goal, minimising dependence on foreign tech is set to become a national security point for every country by 2030.

AI has become a particular flashpoint for this issue, with the US government signalling its willingness to use access to AI models and semiconductor hardware as economic levers to advance its interests.

In response, the EU has invested in developing continental ‘AI gigafactories’, while the UK government has launched a £500 million venture fund to back British AI startups.

FOLLOW US ON SOCIAL MEDIA

Follow ITPro on Google News and add us as a preferred source to keep tabs on all our latest news, analysis, views, and reviews.

You can also follow ITPro on LinkedIn, X, Facebook, and BlueSky.

Adam Shepherd

Adam Shepherd has been a technology journalist since 2015, covering everything from cloud storage and security, to smartphones and servers. Over the course of his career, he’s seen the spread of 5G, the growing ubiquity of wireless devices, and the start of the connected revolution. He’s also been to more trade shows and technology conferences than he cares to count.

Adam is an avid follower of the latest hardware innovations, and he is never happier than when tinkering with complex network configurations, or exploring a new Linux distro. He was also previously a co-host on the ITPro Podcast, where he was often found ranting about his love of strange gadgets, his disdain for Windows Mobile, and everything in between.

You can find Adam tweeting about enterprise technology (or more often bad jokes) @AdamShepherUK.