Qwant’s Flash Answers Now Run on OVHcloud’s Olala Mini, as EUSP’s Staan Index Becomes the Web Layer for a European AI Alliance

The AI answer at the top of Qwant’s results is no longer powered by a model from outside Europe. Qwant’s “réponse flash” (flash answer) now runs on Olala mini 7A1B, the first model from OVHcloud’s in-house lab, served on GPUs from Shadow’s Blade platform. According to Clubic, which covered the partnership announced this morning (Thursday 19 November), it handles three million queries a day.

Qwant is the visible part of a wider deal. Qwant, Shadow, European Search Perspective (EUSP, the joint venture Qwant set up with Ecosia) and OVHcloud have formalised what Clubic calls a technology and commercial partnership to build a European AI stack, running from the web index through to the GPUs.

Who does what

  • EUSP supplies its independent European web index. AI models reach it through Staan, its search API, which offers several endpoints to search pages and retrieve their content for retrieval-augmented generation.
  • OVHcloud supplies the language models from its OVHai LLM lab and the services to run them. Through OVH AI Endpoints, those models will be able to query the web via Staan.
  • Shadow supplies Blade, its GPU platform for training and running models, which Qwant’s search engine already uses.

The two companies will also share model catalogues. Blade users get access to OVHcloud’s open-weight models above 70 billion parameters, the Olala versions and Gladia’s transcription models. Compact open-weight models running on Shadow’s GPUs will be added to OVH AI Endpoints. Clubic says Staan access is also intended for OVH AI Agents, OVHcloud’s agent-building platform, which is due to open in alpha in December 2026.

The deal is between related companies. Clubic notes that Qwant and Shadow both belong to the Synfonium group, whose majority shareholders are Octave and Miroslaw Klaba, the co-founders of OVHcloud.

What Olala mini is

OVH announced Olala on 17 November, ahead of its official launch at today’s OVHcloud AI Summit. Olala mini has 7 billion parameters, of which 1 billion are active, spread across 256 experts per layer. It uses a hybrid attention design (Differential-TPA plus Mamba-3) and has a 32,000-token context window. OVH says it runs on a GPU with as little as 8 GB of memory and reaches 50 tokens per second on a plain CPU. It was trained on three public European supercomputers (Jupiter, MeluXina and Jean Zay) with compute worth about €7m. OVH says it will be released on Hugging Face under Apache 2.0, with weights, the base model, a technical report and AI Act compliance documentation. Version 1.1, due by the end of the year, raises the context window to 128,000 tokens.

For Qwant, the numbers given are answers 20% faster and running costs divided by three compared with the model it used before. OVH doesn’t name that previous model.

Why this matters for search

Staan was already open to third-party search engines and AI providers, and heise reported in July that the French index had “several customers”. Even so, this is a named, commercial deal that puts EUSP’s index under another company’s model platform. When EUSP went live in Germany, the argument was about European search engines relying less on Bing. Now the same index is being sold as grounding for any developer building on OVHcloud’s AI platform. That makes it the European counterpart to the Brave LLM Context API and to DeepSeek’s index in China.

It also says something about how independent engines expect to afford AI answers. Qwant serves a summary on millions of queries a day, and its earlier provider was costing three times as much. A small model that runs on modest hardware, owned by a sister company, is clearly a cost decision as much as a sovereignty one. Ecosia went in a similar direction when it dropped Mistral for open-weight models.

Clubic adds a fair caveat: the offer includes models developed outside Europe, and Olala mini itself was trained on NVIDIA GPUs. Hosting everything in OVHcloud’s European data centres gives control over how the models run, not full technological independence.

My take

For SEOs working in France and Germany, Staan’s crawler and ranking now matter beyond Qwant and Ecosia’s own results pages. If OVH’s agent platform gains traction with European businesses, their internal assistants and agents will cite whatever EUSP indexes and ranks. Check that your logs show EUSP’s crawler reaching your important French and German pages. Watch for OVH AI Agents in December, because that is when Staan grounding moves from a press release to real usage.

One caution: a 1B-active model is small. Flash answers that are fast and cheap but shallow could push more Qwant users to click through to results, which is no bad thing for publishers. I’d like to see Qwant publish quality comparisons, not just speed and cost figures.

Dan Taylor is the founder and editor of International Search Insights (ISI), the publication behind ReadIsi, and a Partner and Head of Innovation (Search & AI) at SALT.agency. He has worked in search for more than a decade, in-house and agency-side, and is regarded as one of the West’s leading voices on search engines beyond Google, including Yandex, Baidu, Naver and Cốc Cốc. On ReadIsi, Dan covers the engines and AI answer platforms that English-language coverage tends to miss: Baidu and ERNIE, Yandex and Alice AI, Naver, Cốc Cốc, Shenma, Sogou, WeChat and Weibo search, plus DeepSeek and Alibaba’s Qwen. He also keeps dated records of documented changes, such as the Yandex, Baidu and Naver algorithm update histories and the DeepSeek, ERNIE and Qwen model update histories. His approach is to take learnings from global search engines and apply logic with a high evidence threshold. A claim needs a source, patterns seen in one market are treated as a hypothesis until they are checked in another, and where the evidence is thin he says so. For brands, that means advice built on what can be shown, with measurement that separates being cited from being chosen. Dan is credited with coining the term “edge SEO” and won the inaugural TechSEO Boost research competition in 2018 for his work on CDN and edge changes for SEO. In 2026 he was credited with coining “Infinite Tail” to describe how keyword research is changing with AI search. He is the author of AI Search Marketing For The Travel Sector, writes for Search Engine Journal, Search Engine Land and Travel Marketing Guy, and has advised technology, SaaS, travel and enterprise ecommerce brands, including Cloudflare, GitLab and Airbnb. He has given multiple talks and panels at the International Search Summit in Barcelona, and has also spoken at Optimization Moscow, brightonSEO, Google Search Central Live Deep Dive Asia in Bangkok, Athens SEO and the Asia SEO Virtual Summit 2026, where his talk looked at how Baidu, Naver, DeepSeek, Tencent, Cốc Cốc and Yandex built their own AI answers.