Introducing Three New Serverless Inference Providers: Hyperbolic, Nebius AI Studio, and Novita 🔥

Source: HuggingFace Date: 2025-02-18 URL: https://huggingface.co/blog/inference-providers-nebius-novita-hyperbolic

Summary

Integration announcement: Hyperbolic, Nebius AI Studio, and Novita join HF Inference Providers, expanding the roster (previously Together AI, SambaNova, Replicate, fal, Fireworks.ai). Notably enables DeepSeek-R1 and FLUX.1 access through the standard HF SDK routing layer. Same economics: PRO users get $2/month credits, no HF markup on routed requests, or bring your own provider API key. No benchmark numbers — this is a routing infrastructure announcement.

Implications

HF as open-source ML hub. Six providers now integrated into the HF inference routing layer. The addition of Hyperbolic and Nebius (both GPU cloud providers with competitive pricing on open-weights models) means the HF router is becoming a meaningful inference marketplace, not just a showcase for a few partners. DeepSeek-R1 availability through all three new providers on the same day as launch reflects coordinated rollout discipline.

Open-weights ecosystem health. The expansion to six providers with no-markup routing creates genuine price competition accessible through a single SDK integration. Teams that previously had to maintain multiple client configurations for cost optimization can now route through HF and let the provider ecosystem compete on price — reducing lock-in risk for any single inference provider.