Nebius Token Factory provides managed inference for open-source language models through an OpenAI-compatible API. Developers can use it for chat completions, reasoning, code generation, embeddings, and related application workloads without managing model-serving infrastructure. Its public serverless endpoints use token-based billing, while optional dedicated endpoints provide reserved capacity and regional deployment controls. The service supports API-key authentication, a browser-based playground, and model-specific performance tiers.

Pricing

Public serverless inference is billed per input and output token, with prices varying by model and performance tier. Free starter credits are promotional and do not constitute an ongoing free plan. Dedicated endpoints use separate capacity-based pricing.

Read more on the pricing page of the service.

Hosting

The core API domain resolves to Nebius B.V. Nebius documents fixed, user-selected deployment regions for dedicated endpoints, including eu-north1 in Finland. Public serverless endpoints are billed per token but can change region; only a fixed EU deployment meets the category's regional hosting condition.