AxForge offering
Subject to confirmed capacity and terms: One NVIDIA RTX 5090 with 32 GB GDDR7 in a dedicated EU-hosted AxForge machine, with full SSH access. Pay per started hour and switch it on and off in the console, or book a week, month or year for a lower rate.
Check the console for current access, prices and availability.
Why AxForge
| Fastest single-stream card on the board | On the same-workload llama.cpp CUDA scoreboard the RTX 5090 decodes at 290.0 t/s with 14,073 t/s prompt processing, roughly 3.8× an RTX 3060. Community measurements, not ours. |
|---|---|
| Modern models at 4k context | Hardware Corner's llama.cpp runs on one RTX 5090: Qwen3 30B-A3B MoE at 226.1 t/s, Qwen3 8B at 200.4 t/s, Qwen3 32B dense at 61.4 t/s. Active parameters and memory traffic beat headline parameter count. |
| One card, by the hour | Pay per started hour and switch the machine on and off in the console, or book a week, month or year for a lower rate. |
Specifications
| Card | One NVIDIA RTX 5090 in a dedicated machine |
|---|---|
| Memory | 32 GB GDDR7 |
| Availability | Bookable in the console: take the next free slot or reserve a start time |
| Tenancy | Dedicated machine, your traffic only, full SSH access |
| Price | €1.26 /hour on demand |
| Longer bookings | from €0.90/hour by the year |
| Region | GPU rentals and bare metal run on our machines in Stockholm (eu-se-1) or Málaga (eu-es-1); the console shows each machine's city. AxForge serverless inference runs in Stockholm. |
| Community decode | 290.0 t/s single-card (llama.cpp CUDA scoreboard, Llama 2 7B Q4_0). Labelled, not ours |
RTX 5090 catalogue reference: specifications and requirements. List prices in EUR.
How it works
| 1 | Confirm capacity and terms in the console before the rental steps below. A public request records demand; it does not reserve or provision a machine. |
|---|---|
| 2 | Sign in to AxForge. |
| 3 | Choose RTX 5090 and the term: on demand, week, month or year. |
| 4 | Pay in the console; the machine is prepared for you. |
| 5 | Add your SSH key and connect when the machine is yours. |
| 6 | End the rental when your workload is complete. |
FAQ
Capacity and terms must be confirmed before a booking. Check availability in the console and take the next free slot, or reserve a start time.
One NVIDIA RTX 5090 with 32 GB GDDR7 in a dedicated machine serving only your traffic, with full SSH access.
Capacity and terms must be confirmed before a booking. Listed per hour on demand; a week, month or year earns a lower rate. The prices are on this page and in the console.
Models and MoEs that fit in 32 GB. It is the fastest-per-euro choice for them. When a model no longer fits in VRAM and starts spilling, decode collapses; for larger single-GPU models look at the 96 GB RTX 6000 Pro or the 128 GB DGX Spark.
GPU rentals and bare metal run on our machines in Stockholm (eu-se-1) or Málaga (eu-es-1); the console shows each machine's city. AxForge serverless inference runs in Stockholm.
No. Your model, your traffic, our hardware. Prompts are never persisted. Only operational metadata is kept for billing and operations.