inference/research
verifiedhigh · 0.85updated 2026-10-16

Nebius Token Factory managed inference

other

Fields

id
services/svc-nebius-token-factory
Provider
other
Endpoint
https://api.tokenfactory.nebius.com/v1
Input $/Mtok
0.15
Output $/Mtok
0.6
Price basis
USD per 1M TOKENS, representative model openai/gpt-oss-120b at $0.15 in / $0.60 out, context 131k, region eu-north1, from the public Token Factory catalog fetched 2026-10-16. Same-day siblings on that catalog: Qwen3-235B-A22B-Instruct-2507 $0.20/$0.60 (262k, eu-north1), Qwen3-30B-A3B-Instruct-2507 $0.10/$0.30 (262k), gemma-3-27b-it $0.10/$0.30 (110k), DeepSeek-V4-Flash-0731 $0.14/$0.28 (us-central1). The 2026-10-14 record left both price fields null because only the authenticated console was reachable then; the public catalog.md now publishes the rate card, so this replaces that null rather than contesting it. No Llama 3.3 70B row was on the catalog view fetched today, so that slug was dropped from models_served.
Context window
131,072
Regions
eu-north1eu-west2uk-south1us-north1us-central1
notes
Price is now first-party: tokenfactory.nebius.com/model-catalog.md, which the page itself describes as the reading view of /api/public/models_info. The 2026-10-14 source (svc-src-nebius-token-factory-2026-10-14) still documents why the price was null — the public nebius.com/prices page is GPU-hour pricing (HGX H100 $4.50, H200 $5.40, B200 $8.50) and the token card was console-gated. models_served lists only slugs that exist in data/models/ and appeared on the 2026-10-16 catalog; the catalog is much larger (Kimi, GLM, Nemotron, Hermes-4-405B). No third-party TTFT/tok-s for Token Factory was on the Artificial Analysis Llama 3.3 70B provider page (Nebius is not in that set), so measured_* stay null. Endpoint remains the OpenAI-compatible base https://api.tokenfactory.nebius.com/v1.

Notes

Price is now first-party: tokenfactory.nebius.com/model-catalog.md, which the page itself describes as the reading view of /api/public/models_info. The 2026-10-14 source (svc-src-nebius-token-factory-2026-10-14) still documents why the price was null — the public nebius.com/prices page is GPU-hour pricing (HGX H100 $4.50, H200 $5.40, B200 $8.50) and the token card was console-gated. models_served lists only slugs that exist in data/models/ and appeared on the 2026-10-16 catalog; the catalog is much larger (Kimi, GLM, Nemotron, Hermes-4-405B). No third-party TTFT/tok-s for Token Factory was on the Artificial Analysis Llama 3.3 70B provider page (Nebius is not in that set), so measured_* stay null. Endpoint remains the OpenAI-compatible base https://api.tokenfactory.nebius.com/v1.

Cross-references

Models (5)