AxForge
Sign in Get started

GPU rental · dedicated hardware

Rent a dedicated NVIDIA DGX Spark (GB10)

Your own 128 GB GB10 Grace-Blackwell machine with full SSH access, hosted by AxForge in Málaga, Spain (eu-es-1). Rent it by the hour, week, month or year. Take the next available capacity or reserve a start time in advance. Your model, your traffic, your machine.

eu-es-1 · Málaga Available Dedicated · full SSH access from €0.85/hour dgx-spark.axforge.ai
Check availability Reserve capacity View benchmarks €1.19/hour on demand · €0.95/hour by the week · €0.91/hour by the month · €0.85/hour by the year. Choose the rental period that fits your workload. Reserve longer sessions in advance or take the next available capacity.
NVIDIA DGX Spark

Specifications

What you rent

SystemNVIDIA DGX Spark, dedicated, single-tenant
SuperchipNVIDIA GB10 Grace-Blackwell, see the GB10 page
Memory128 GB unified, shared CPU/GPU
CPU architectureARM64
AccessFull SSH access: your own environment, containers and serving stack
TenancyDedicated: the machine is yours for the duration of the rental, not shared capacity
Rental termHour, week, month or year
Pricing€1.19/hour on demand · €0.95/hour by the week · €0.91/hour by the month · €0.85/hour by the year
AvailabilityCheck current availability in the AxForge Console or reserve the next available start time.
RegionMálaga, Spain (eu-es-1)
TrialTrial access is available for qualifying customers.

Performance

What this machine does, by workload

Example workloadModelCommunity benchmarks*
Fast agents / reasoning (MoE)Qwen3.6 35B-A3B · NVFP4 + MTPup to 86.3 t/s
Coding (MoE)Qwen3-Coder 30B-A3B~42 to 61 t/s
Dense reasoningQwen3.8 27B · NVFP4 + MTP25.1 t/s
Smaller dense LLMQwen3 8B · Q4~42 t/s
Large MoEDeepSeek V4 Flash · 180B, 13B active~23 t/s

*Community results (SparkBench PBM @ 4k, llama.cpp), labelled, not ours; they vary by runtime, quantization, context length and serving configuration. Full tables, sources and the MoE-vs-dense story on the GB10 page. The same hardware also serves image generation and editing. See the model catalogue.

Which models fit its memory (about 100 GB of the 128 GB is for the model): language models · image models · video models, estimated from each model's own files and config.

Fit

What a DGX Spark is for

Use caseWhy it fits
Dedicated inference, ~7B to 35B models128 GB unified memory holds model weights and KV cache in one pool, and the community numbers above are exactly this workload.
Private inferenceA single-tenant machine in an EU region. Your model, your traffic, our hardware. Prompts are never persisted.
Dev / staging nodesA machine you keep for the day, week or month, a stable target for integration, load testing and pre-production serving.
Model evaluationRun bake-offs on the exact hardware class you would serve from, with results that transfer.
AI agentsHost agent runtimes next to their model, long-running processes with no per-token surprise from a third party.
Fine-tuning experimentsThe unified memory pool fits adapters and smaller fine-tunes that would need careful sharding elsewhere.
Self-managed serving & containersBring your own serving stack, like vLLM, llama.cpp or custom containers, and run it your way on your machine.

Searching for DGX Spark cloud or DGX Spark hosting? Same machine: you rent a dedicated physical system, we host and operate it, you reach it over the network.

ARM64

It's an ARM64 machine. Here's what that means

The GB10 platform is ARM64, not x86. Most modern AI frameworks and container images ship ARM64 builds, and our own production serving stack runs on it. x86-only binaries need ARM64 builds or rebuilds. Not sure about your stack? An engineer can review it with you before you commit.

How it works

From sign-in to running workload

1Sign in to AxForge.
2Choose your rental duration. Hour, week, month or year.
3Check or reserve available capacity. Take the next available machine or schedule a start time.
4Add your SSH key.
5Connect to your dedicated machine and deploy your models, containers and tools.
6End the rental when your workload is complete.

Check availability or Reserve capacity in the console. Trial access is available for qualifying customers.

Data & privacy

Your model, your traffic, your machine

Prompts never persisted. Requests to your dedicated DGX Spark are processed on your own machine in Málaga, Spain (eu-es-1), not logged, not retained and never used by AxForge to train anything. We keep only operational metadata (rental period, machine status) for billing and operations. Full policy at axforge.ai/privacy.

FAQ

DGX Spark rental: common questions

Can I rent a DGX Spark today?

Check the live availability in the console. Immediate and scheduled capacity are both supported. Check availability.

How does provisioning work?

Reserve capacity, add your SSH key and receive access to your allocated system for the booked period. From there you deploy your own models, containers and tools, and end the rental when your workload is complete.

How much does DGX Spark rental cost?

€1.19/hour on demand · €0.95/hour by the week · €0.91/hour by the month · €0.85/hour by the year for a dedicated machine. Choose the rental period that fits your workload. Longer reservations receive a lower effective compute rate. Full list on the pricing page.

What workloads can I run on a rented DGX Spark?

Anything that fits a 128 GB ARM64 machine you fully control: dedicated inference of ~7B to 35B models, agents, fine-tuning experiments, evaluation, dev/staging, and your own serving stack in containers. If you only need per-token inference, the serverless Qwen API may be enough, no machine required.

What models run well on a DGX Spark?

Dense models of roughly 7B to 35B, and MoE models far larger: community benchmarks run 80B to 180B MoE at usable speeds, because only the active experts are read per token. Full tables and sources on the GB10 page.

Is DGX Spark hosting the same as DGX Spark cloud?

On AxForge you rent a dedicated physical machine, hosted and operated by us in Málaga, Spain (eu-es-1), reachable over the network like any cloud endpoint, but it is your dedicated system, not a shared cloud instance.

DGX Spark is ARM64. Will my stack run on it?

The GB10 platform is ARM64. Most modern AI frameworks and container images ship ARM64 builds, and our own serving stack runs on it in production. x86-only binaries need ARM64 builds; an engineer can review your stack before you commit.

Is trial access available?

Trial access is available for qualifying customers. Request trial access from the console.

Are my prompts stored on a rented DGX Spark?

No. Your model, your traffic, your dedicated machine. Prompts are never persisted by AxForge. Only operational metadata is kept for billing and operations. See the privacy policy.

Ready to run on your own DGX Spark?

Check availability Talk to an engineer

Explore

All GPUs

dgx-spark.axforge.ai, the short link for this product → GB10 deep-dive & benchmarks RTX 3060: one card or two, €0.24/hour Qwen3.8 27B on DGX Spark RTX 6000 Pro: request capacity RTX 5090: request capacity RTX 3090: request capacity H100: request capacity H200: request capacity B200: request capacity DGX Station: request capacity All GPUs in the catalogue Just need the API? Qwen3.8 27B → Hosted models → Pricing →
© 2026 AxForge · EU-hosted AI infrastructure Pricing Docs Trust Privacy Terms