OVHcloud launches AI Endpoints for serverless model access

OVHcloud has introduced AI Endpoints, a serverless solution providing developers with access to over 40 open-source large language and generative AI models for application development.

The newly launched platform is designed to enable integration of functions such as chatbots, text-to-speech and coding assistance into software, without requiring expertise in machine learning or management of underlying infrastructure. Models are hosted in a cloud environment with an emphasis on data protection.

Yaniv Fdida, Chief Product and Technology Officer at OVHcloud, highlighted the scope of AI Endpoints and the company's ongoing engagement with the developer community. Fdida stated, "We are excited to launch AI Endpoints and are humbled by the incredible feedback we get from our amazing community. With support for the most diverse and sought after open source LLM models, AI Endpoints helps to democratise AI so developers can add to their apps the most cutting-edge models. Our solution enables them to do this easily in a trusted cloud environment with full confidence in OVHcloud's sovereign infrastructure."

AI Endpoints is designed to facilitate the testing and deployment of AI features, allowing developers to experiment with models in a sandbox environment before integrating them into production environments or business processes. Example use cases include deploying conversational agents, extracting text from documents, enabling speech-to-text capabilities, and delivering real-time coding assistance within development environments.

The service addresses a range of applications, with its models supporting real-time natural language processing, automating customer service tasks, and analysing unstructured data for business processes. Coders can utilise features such as private, real-time AI-based code suggestions and error detection tools to improve productivity and code quality.

OVHcloud emphasises its sovereign cloud infrastructure, with all data managed in European data centres. According to the company, this approach ensures that developers know their information is hosted within Europe and protected from non-European regulatory requirements, delivering both technical and strategic autonomy. The AI Endpoints platform is powered by energy-efficient infrastructure, with water-cooled servers located in environmentally certified data centres, intended to mitigate the environmental impact of intensive AI workloads.

With transparency as a key principle, AI Endpoints uses open weight models, which, according to OVHcloud, allows organisations to retain control over their data and maintain flexibility regarding the deployment of models on-premises or across other cloud services.

The available suite of models includes Llama 3.3 70B, Mixtral 8x7B, Mistral Nemo, Qwen 2.5 Coder 32B, Codestral Mamba, DeepSeek-R1, Qwen 2.5 VL 72B, SDXL for image generation, and support for speech-to-text (ASR) and text-to-speech (TTS) functions. The service supports a pay-as-you-go pricing structure, varying by model and based on the number of tokens processed per minute.

Having completed a customer feedback-led development and testing phase, AI Endpoints has incorporated additional options for model selection, enhanced API key management, and support for stable open-source models. The service is offered in the Asia-Pacific, Canadian, and European markets and is currently deployed from OVHcloud's Gravelines data centre.