Nebius vs Replicate
Same workloads, both price lists, refreshed daily. On shared line items today: Nebius is cheaper on 3 of 5 shared models (input or output price), Replicate on 0.
Nebius
AI cloud from the former Yandex team with H100/H200/B200 clusters in Europe and the US, plus Nebius AI Studio for per-token open-weight inference. Competitive pricing at cluster scale.
- European datacenters
- Both clusters and per-token inference
- Aggressive large-scale pricing
Replicate
Run thousands of community models (LLMs, image, audio, video) behind one API with per-second GPU billing: the fastest path from 'found a model on GitHub' to production endpoint..
- Massive community model zoo
- Per-second billing, scale to zero
- Custom models via Cog
Shared models, priced by both
List prices refreshed 2026-11-22 ยท input $/1M tokens| Model | Nebius in/out | Replicate in/out | Cheaper input |
|---|---|---|---|
| GPT-OSS 120B | $0.150 / $0.600 | $0.180 / $0.720 | Nebius |
| Qwen3 235B A22B | $0.200 / $0.600 | $0.264 / $1.06 | Nebius |
| DeepSeek V3 | $0.500 / $1.50 | $1.45 / $1.45 | Nebius |
| DeepSeek R1 | $0.800 / $2.40 | $3.75 / $10.00 | Nebius |
| Kimi K2.6 | $0.950 / $4.00 | $0.950 / $4.00 | tie |
Cheaper-on-count is a tally of listed prices, not a quality verdict: throughput, reliability and quantization differ between providers. Disclosure.