Inference Router (Public Preview)
A policy-aware control plane that routes every request based on cost, latency, quality, and residency constraints. Instead of hardcoding model logic, teams define intent once and the system optimizes execution across providers and deployment types.
Dedicated Inference and Bring Your Own Model
Run custom and fine-tuned models on dedicated GPU infrastructure with full control over performance, scaling, and configuration. Deploy models from sources like Hugging Face or your own environments, and operate high-throughput workloads with a pre-tuned inference stack and managed orchestration, without the complexity of Kubernetes.
Expanded Models and Services
Run and evaluate text, image, audio, and video models through a single system. A continuously refreshed catalog includes 25+ new models and Day 0 access to select releases, including NVIDIA Nemotron 3 Nano Omni, available first on DigitalOcean. The highly efficient open multimodal model unifies vision, speech, language, and tool use, with NVIDIA TensorRT-LLM tuned at the kernel level. Built-in evaluations let teams benchmark quality, cost, and latency before production.
PostgreSQL & MySQL Advanced Edition (Public Preview)
Managed PostgreSQL and MySQL Advanced Edition deliver hyperscaler-grade reliability and scale, available alongside Standard Edition.
Managed Weaviate (Private Preview)
Production-ready vector infrastructure without the operational overhead, with native integration to Serverless Inference and predictable pricing. Sign up for early access.
Knowledge Bases
A fully managed RAG service that handles ingestion, chunking, embedding, retrieval, and reranking, with MCP support for agent frameworks. Customers are moving from prototype to production in days.
New feature available: Configurable Contacts
You can now control who receives critical account notifications with our newest Teams feature, Configurable Contacts. We’ve enabled billing contacts for invoices, receipts, and payment issues and security contacts for login alerts and risk events. With this feature, alerts go directly to the right users without manual forwarding. You can get started by going to Settings (on the Teams page) → Contacts → Add billing or security contact.
