Estimated 500-token answer. Across 12 models with recent complete speed data.
Average response start
1.31 s
Wait until the first token. Across 12 models with recent response start data.
Average output speed
39.7 tok/s
Generated tokens per second. Across 12 models with recent output speed data.
AI Models available from Nebius
Prices and recent speed measurements below are for Nebius endpoints. A 500-token response combines the wait for the first token with the time to generate 500 output tokens.
How to evaluate Nebius as an AI inference provider
Nebius currently has 13 indexed models on this site, including 12 text, 1 embeddings. This page combines the provider-specific catalog with published pricing and recent performance data so you can identify suitable models and then compare Nebius with alternative hosts on each model page.
Review model coverage
Nebius offers model-specific routes rather than one interchangeable service. Check modality, context window, supported inputs, and parameter support before comparing price. The largest published context window in this catalog is 1M tokens.
Interpret price and speed together
The lowest published input price shown here is $0.01 / 1M tokens. Recent 500-token response measurements are available for 12 models; missing measurements remain visible and are never estimated.
Confirm deployment requirements
Nebius lists its headquarters as NL. Published datacenter coverage: no datacenter regions are currently published. Headquarters and hosting regions describe different things, so verify data residency, service terms, and current availability directly with the provider.
