Fine-tuning and inference for large LLMs
Whole node — $330,000 · on-site hardware audit
SLA and a dedicated manager
Hours shared across the team
8× H100 · 640 GB HBM · NVLink
per 1/200 share · ≈28 GPU-h/mo
Templates: chat LLM, Flux, ComfyUI, API
Whole server — $3,600: SSH, your own models, rent
it out entirely
Resell the share on the secondary market in
one click
Private encrypted volume, 100 GB
RTX 4090 · 24 GB · Irkutsk data center
per 1/12 share · ≈60 h/mo
Everything in Start, plus video models
Whole server — $12,000: SSH, your own models,
rent it out entirely
Multi-session: 3 GPUs in parallel
250 GB volume + priority in the pool
3× RTX 5090 · 96 GB total · Bratsk
per 1/16 share · ≈135 GPU-h/mo