Brands and logos belong to their respective owners and are shown only to indicate compatibility. Registro Digital is not affiliated with them. Each model has its own usage license.
Light inference
GPU T4
7 to 8 billion parameter models, audio transcription and vision.
- GPU
- 1 × NVIDIA Tesla T4
- Video memory
- 16 GB
- Processor
- AMD EPYC 7302P · 16 cores
- RAM
- 128 GB DDR4 ECC
- Storage
- 2 × 1 TB NVMe (RAID 1)
$997.10USD per month · excl. VAT
One-time setup: $997.10 USD excl. VAT
Request this server
Most popular
GPU A10
Production chatbots with compressed models of up to 27 billion parameters.
- GPU
- 1 × NVIDIA A10
- Video memory
- 24 GB
- Processor
- AMD EPYC 7313P · 16 cores
- RAM
- 128 GB DDR4 ECC
- Storage
- 2 × 960 GB NVMe
$1,267.50USD per month · excl. VAT
One-time setup: $1,267.50 USD excl. VAT
Request this server
Video and inference
GPU Flex 170
Video transcoding and inference with OpenVINO. Does not use CUDA.
- GPU
- 1 × Intel Data Center GPU Flex 170
- Video memory
- 16 GB
- Processor
- Intel Xeon Gold 5412U · 24 cores
- RAM
- 256 GB DDR5 ECC
- Storage
- 2 × 1.92 TB NVMe
$1,335.10USD per month · excl. VAT
One-time setup: $1,335.10 USD excl. VAT
Request this server
Large models
GPU RTX PRO 6000
30 to 120 billion parameter models, images, video and fine-tuning.
- GPU
- 1 × NVIDIA RTX PRO 6000 Blackwell
- Video memory
- 96 GB
- Processor
- 2 × Intel Xeon 6517P
- RAM
- 256 GB
- Storage
- 2 × 1.92 TB NVMe
$3,042.00USD per month · excl. VAT
One-time setup: $3,042.00 USD excl. VAT
Request this server
Maximum performance
GPU H200 S
70 billion parameter models with many concurrent users.
Ideal for
vLLM
High concurrency
Llama 3.3 70B
gpt-oss-120b
Qwen3 32B
Fine-tuning
Models of up to 30 billion
Qwen3 32B
- GPU
- 1 × NVIDIA H200
- Video memory
- 141 GB
- Processor
- 15 dedicated vCPUs
- RAM
- 267 GB
- Storage
- 1 TB
- Location
- Germany (EU)
$3,094.69USD per month · excl. VAT
No setup fee
Request this server
2 GPUs
GPU H200 M
Models with more than 100 billion parameters.
Ideal for
vLLM
Models with several hundred billion parameters
Qwen3 235B
MiMo-V2-Flash
Llama 3.3 70B
Fine-tuning
LoRA on 70 billion parameter models
Llama 3.3 70B
- GPU
- 2 × NVIDIA H200
- Video memory
- 282 GB (141 GB per GPU)
- Processor
- 30 dedicated vCPUs
- RAM
- 534 GB
- Storage
- 1.5 TB
- Location
- Germany (EU)
$6,188.43USD per month · excl. VAT
No setup fee
Request this server
4 GPUs
GPU H200 L
The largest open models, compressed, and training.
Ideal for
vLLM
Compressed frontier models
GLM-5.2
DeepSeek V3
Qwen3 235B
Training
Distributed training on 4 GPUs
- GPU
- 4 × NVIDIA H200
- Video memory
- 564 GB (141 GB per GPU)
- Processor
- 60 dedicated vCPUs
- RAM
- 1068 GB
- Storage
- 2 TB
- Location
- Germany (EU)
$12,377.81USD per month · excl. VAT
No setup fee
Request this server
8 GPUs
GPU H200 XL
The largest open models at maximum quality, and training.
Ideal for
vLLM
Frontier models in 8-bit
GLM-5.2
DeepSeek V3
Kimi K2
Training
Custom models and multiple models in parallel
- GPU
- 8 × NVIDIA H200
- Video memory
- 1128 GB (141 GB per GPU)
- Processor
- 127 dedicated vCPUs
- RAM
- 2136 GB
- Storage
- 4 TB
- Location
- Germany (EU)
$24,755.61USD per month · excl. VAT
No setup fee
Request this server
GPU servers are activated subject to availability; we confirm the delivery date before payment. They include 2 TB of outbound traffic and 1 public IPv4.