How they use Nebius AI Cloud
FriendliAI helps startups and enterprises deploy and serve AI models through inference APIs and dedicated endpoints. Its APIs support applications such as customer support bots, coding assistants and AI agents. FriendliAI runs its inference stack and APIs on Nebius GPU infrastructure, combining its proprietary inference optimization technology with Nebius compute.
Its November 2025 announcement describes adding Nebius to its multicloud capacity. FriendliAI says the combined offering automatically adjusts GPU resources to inference demand and guarantees GPU availability for sudden spikes or large workloads. It advertises over 50% lower GPU costs, more than twice the inference speed and a 99.99% uptime guarantee for the combined offering.
Commercial use. FriendliAI runs customer-facing inference APIs on Nebius infrastructure.
Sources
2 sourcesFriendliAI Expands Ultra-Fast AI Inference Platform with Nebius AI Cloud Integration
Listed 21 Oct 2026 · The National Law Review · Web page
FriendliAI’s syndicated press release describes its APIs running on Nebius infrastructure for inference. Its title names Nebius AI Cloud Integration. The publication header states October 28, 2025.
Open original sourceCompany context
friendli.aiFriendliAI builds infrastructure for running open-weight and custom AI models. Developers and businesses use its Model APIs for managed inference, Dedicated Endpoints for dedicated model serving, and BYOG to run FriendliAI's inference stack on their own cloud GPUs. The platform handles deployment, scaling and inference optimization for applications such as coding assistants, AI agents and chatbots. In its October 2025 Nebius announcement, FriendliAI reported serving trillions of tokens a month across its platform.
Recent updates
6 updatesFriendliAI joins Vercel AI Gateway
FriendliAI announced that developers can access its open-weight model inference through Vercel AI Gateway. The integration supports provider fallback routing and payment through Gateway credits or a customer's FriendliAI API key.
FriendliAIFriendliAI adds NVIDIA Nemotron Retriever models to Dedicated Endpoints
FriendliAI made NVIDIA Nemotron 3.5 Embed 1B, Embed 8B and Rerank 8B available on Dedicated Endpoints. The models support retrieval across text and visual content for enterprise search and AI agents.
FriendliAIFriendliAI makes NVIDIA Nemotron 3.5 Super VL available
FriendliAI added NVIDIA Nemotron 3.5 Super VL to Dedicated Endpoints. The multimodal model combines visual understanding, reasoning and tool use for workloads involving text, images, documents and video.
FriendliAI