Provider pricing

Nebius AI Studio AI API pricing

Nebius AI Studio (spun out of Yandex's cloud business) offers serverless and dedicated inference for Llama 4, DeepSeek V3.2, Qwen 3, and Mixtral on its own H100/H200 GPU fleet across Finland, France, and the US. Pricing is per-million-tokens with sub-cent rates on smaller models, plus on-demand GPU rental.

Founded2023 (rebrand from Yandex N.V.)
HeadquartersAmsterdam, Netherlands
Tracked modelsDirectory

Best for

Cost-sensitive teams running open models, especially in Europe.

Strengths

  • Aggressive per-token pricing — often the cheapest serverless tier
  • EU data residency (Finland, France) for GDPR-sensitive workloads
  • Owns the underlying H100/H200 fleet — no resale markup
  • On-demand GPU rental in addition to serverless inference