High-performance GPUs on Scaleway's cloud to accelerate your model training and inference cost-effectively.
Get started with €100 free credit when you create a Business account.
Access NVIDIA's most powerful cloud GPU Instances and push the boundaries of performance for your AI models. These powerhouses leverage NVLink, NVIDIA's low-latency GPU-to-GPU interconnect, delivering the massive bandwidth and multi-node scalability required for the training and inference of advanced Large Language Models (LLMs).
Ideal for superior LLM performance and real-time reasoning.
From €7.5/GPU/hour¹
Billed per minute
¹ Price and specs for 8 GPUs
Ideal for LLM fine-tuning and larger LLM model inference.
From €2.88/GPU/hour¹
Billed per minute
¹ Price and specs for 8 GPUs
Scale your AI production efficiently with our versatile GPU lineup. Whether you need cost-effective LLM fine-tuning for 7B-70B models, high throughput GenAI or low-latency inference, these GPU Instances offer the perfect balance of performance and price. Drive value across a wide range of workloads without breaking the bank.
Billed per minute
Billed per minute
Billed per minute
Generate new text leveraging the main language models such as base (e.g., Mistral 7B), instruction-trained (e.g., Llama-3.1-70B-Instruct), human feedback, and reasoning (e.g., Deepseek R1).
Beyond classical computer vision (classification, segmentation...), analyze images for tasks such as Optical Character Recognition (OCR) and other computer vision applications beyond character recognition, with models such as pixtral-12B-2409.
Transform text into natural-sounding speech, and vice-versa with models such as kyutai/moshika-0.1-8b (speech-to-speech) and WhisperX (text-to-speech and speech-to-text). Speech generation models enable a wide range of applications, including accessibility tools and multilingual communication.
Generate, complete, and debug code automatically with models such as qwen2.5-coder-32b-instruct. These models are trained on vast datasets of programming languages and can understand instructions or context to create code snippets, solve coding challenges, and even explain code.
| GPU | VRAM | RAM | Performance | Recommended for | Price per GPU | |
|---|---|---|---|---|---|---|
| 8x NVIDIA B300-SXM | 8x80 GB | 3,840 GB DDR5 | 8x1,979 TFLOPS | Large scale AI model training & inference workloads | €60 /Hour | |
| 8x NVIDIA H100-SXM | 8x80 GB | 960 GB DDR5 | 8x1,979 TFLOPS | LLM fine-tuning & inference | €23.03 /Hour | |
| 4x NVIDIA H100-SXM | 4x80 GB | 480 GB DDR5 | 4x1,979 TFLOPS | LLM fine-tuning & inference | €11.61 /Hour | |
| 2x NVIDIA H100-SXM | 2x80 GB | 240 GB DDR5 | 2x1,979 TFLOPS | LLM fine-tuning & inference | €6.02 /Hour | |
| 1x NVIDIA H100 | 80 GB | 240 GB DDR5 | 1,513 TFLOPS | 7B LLM fine-tuning & inference | €2.73 /Hour | |
| 8x NVIDIA L40S | 8x 48 GB | 768 GB DDR4 | 2,896 TFLOPS | 70B text-to-image model fine-tuning & inference | €11.20 /HOUR | |
| 4x NVIDIA L40S | 4x 48 GB | 384 GB DDR4 | 1,448 TFLOPS | 7B text-to-image model fine-tuning & inference | €5.60 /HOUR | |
| 2x NVIDIA L40S | 2x 48 GB | 192 GB DDR4 | 724 TFLOPS | GenAI (image/video) | €2.80 /HOUR | |
| 1x NVIDIA L40S | 1x 48 GB | 96 GB DDR 4 | 362 TFLOPS | GenAI (image/video) | €1.4 /HOUR | |
| 8x NVIDIA L4 | 8x 24 GB | 384 GB DDR4 | 8x 242 TFLOPS | 70B LLM Inference | €6.00 /HOUR | |
| 4x NVIDIA L4 | 4x 24 GB | 192 GB DDR4 | 4x 242 TFLOPS | 7B LLM Inference | €3.00 /HOUR | |
| 2x NVIDIA L4 | 2X 24 GB | 96 GB DDR4 | 2x 242 TFLOPS | Video Encoding (8K) | €1.50 /HOUR | |
| 1x NVIDIA L4 | 1x 24 GB | 24 GB DDR4 | 242 TFLOPS | Image Encoding (8K) | €0.75 /HOUR |
The most complete cloud ecosystem in Europe, from Bare Metal to serverless and everything in between.
Our wide range of services is designed to sustain your growth cost-effectively, at all stages of development.
Our products are compatible with market standards so that you can enjoy the freedom of no lock-in.
100% of electricity consumed in our data centers comes from renewable energy. Decommissioned hardware is securely reused & recycled.
