Skip to main content

SAFOZi

GPU as a service

The Most Powerful Servers in Africa

● GPU CLOUD — ALL SYSTEMS ONLINE ● CLOUD GPU — TOUS SYSTÈMES OPÉRATIONNELS

GPU + Models + Storage + Metrics GPU + Modèles + Stockage + Métriques

Bare-metal NVIDIA GPUs, one-click hosted models, persistent storage, and real-time metrics — everything to run AI at scale. GPUs NVIDIA bare-metal, modèles hébergés en un clic, stockage persistant et métriques en temps réel — tout pour faire tourner l’IA à grande échelle.

RTX PRO 6000 · B200Available GPUsGPUs disponibles
<60sProvisioningProvisionnement
99.99%Uptime SLASLA Uptime
GPU CATALOGCATALOGUE GPU

Cloud GPUsGPUs Cloud

NVIDIA RTX PRO 6000 Blackwell
Versatile · Dev · Image Gen
96GB GDDR7VRAM
1.8 TB/sBANDWIDTH
300WTDP
DynamicProvisioning
$1.32 / hour/ heure
Available (US-WEST-1/2/3)
NVIDIA B200 Blackwell
Flagship · LLM Training · Inference
180GB HBM3eVRAM
8 TB/sBANDWIDTH
1000WTDP
NVLink 5Multi-GPU
$7.50 / hour/ heure
Available (LA-1/2/3)
NVIDIA L40S
Balanced · Rendering · Inference
48GB GDDR6VRAM
864 GB/sBANDWIDTH
350WTDP
8 vCPUCompute
$1.84 / hour/ heure
Unavailable (À annoncer / TBA)
NVIDIA RTX 4090
Budget · Dev · Inference
24GB GDDR6XVRAM
1 TB/sBANDWIDTH
450WTDP
AdaArchitecture
$0.78 / hour/ heure
Unavailable (À annoncer / TBA)
APPSAPPLICATIONS

Pre-Configured ApplicationsApplications Pré-Configurées

⚗
Jupyter + PyTorch
development · ~5 min
Full Python development environment with PyTorch, CUDA, and JupyterLab
Deploy:Déployer: ~5 min Min VRAM:Min VRAM: 4GB
⌨
VS Code Server
development · ~3 min
Full VS Code IDE in your browser with GPU access
Deploy:Déployer: ~3 min Min VRAM:Min VRAM: 4GB
⛓
Langflow
development · ~5 min
Visual framework for building LLM applications
Deploy:Déployer: ~5 min Min VRAM:Min VRAM: 4GB
⏱
MLflow
development · ~3 min
ML experiment tracking and model registry
Deploy:Déployer: ~3 min Min VRAM:Min VRAM: 4GB
V1
✦
vLLM V1 + TinyLlama
inference · ~5 min
One-click vLLM V1 with TinyLlama 1.1B — ready for immediate inference
Deploy:Déployer: ~5 min Min VRAM:Min VRAM: 4GB
FAST
⚡
vLLM Inference Server
inference · ~8 min
High-performance LLM inference with V1 engine and OpenAI-compatible API
Deploy:Déployer: ~8 min Min VRAM:Min VRAM: 16GB
✍
Text Generation WebUI
inference · ~12 min
Feature-rich web interface for running LLMs (oobabooga)
Deploy:Déployer: ~12 min Min VRAM:Min VRAM: 8GB
NEW
❝
Open WebUI
inference · ~3 min
ChatGPT-like interface for local LLMs (works with Ollama)
Deploy:Déployer: ~3 min Min VRAM:Min VRAM: 4GB
☺
Hugging Face TGI
inference · ~10 min
Production-ready text generation inference from Hugging Face
Deploy:Déployer: ~10 min Min VRAM:Min VRAM: 16GB
♆
Triton Inference Server
inference · ~8 min
NVIDIA’s production inference server for any ML framework
Deploy:Déployer: ~8 min Min VRAM:Min VRAM: 8GB
⌂
LocalAI
inference · ~5 min
Drop-in OpenAI API replacement for local inference
Deploy:Déployer: ~5 min Min VRAM:Min VRAM: 4GB
CREATIVE
⬡
ComfyUI
creative · ~10 min
Node-based Stable Diffusion interface for image generation
Deploy:Déployer: ~10 min Min VRAM:Min VRAM: 8GB
✏
Automatic1111 WebUI
creative · ~15 min
The most popular Stable Diffusion web interface
Deploy:Déployer: ~15 min Min VRAM:Min VRAM: 8GB
EASY
◎
Fooocus
creative · ~10 min
Simplified Stable Diffusion with Midjourney-like UX
Deploy:Déployer: ~10 min Min VRAM:Min VRAM: 8GB
NEW
▶
CogVideoX
creative · ~15 min
State-of-the-art AI video generation
Deploy:Déployer: ~15 min Min VRAM:Min VRAM: 24GB
⚒
Axolotl Fine-tuning
training · ~8 min
Fine-tune LLMs with LoRA, QLoRA, and full fine-tuning
Deploy:Déployer: ~8 min Min VRAM:Min VRAM: 24GB
⚙
Kohya_ss GUI
training · ~12 min
LoRA and fine-tuning for Stable Diffusion models
Deploy:Déployer: ~12 min Min VRAM:Min VRAM: 12GB
PLATFORMPLATEFORME

Everything You NeedTout ce qu'il vous faut

▤

Persistent Storage

High-speed NVMe volumes that survive instance restarts. Mount across multiple GPUs simultaneously.

  • NVMe-backed block storage
  • Up to 10TB per volume
  • Cross-region replication
  • Snapshot & restore
  • S3-compatible object store
  • Tarifs sur demande
❖

Hosted Models

Deploy and serve open-source models via a managed inference API — no GPU management required.

  • OpenAI-compatible REST API
  • Llama 3, Mistral, Qwen, Phi
  • Stable Diffusion / FLUX
  • Whisper, CLIP, Embeddings
  • Auto-scaling by demand
  • Deployed from your GPU credit
↗

Metrics & Monitoring

Real-time GPU telemetry, cost tracking, and usage analytics for every instance you run.

  • GPU utilization & VRAM usage
  • Temperature & power draw
  • Network I/O per instance
  • Cost forecasting dashboard
  • Prometheus-compatible export
  • Alerts & anomaly detection
PRICINGTARIFICATION

Simple, Transparent PricingTarification Simple et Transparente

Pay per Use

GPU Credit

$1.32/h
Starting price — deducted from your prepaid GPU credit
  • RTX PRO 6000 — $1.32/h
  • B200 — $7.50/h
  • L40S & RTX 4090 — coming soon
  • Pay only for the hours you use
  • All pre-configured apps included
  • Hosted models via API
Monthly

Subscription

$598/mo
Starting price — fixed monthly price, billed every month
  • RTX PRO 6000 Blackwell (96 GB) — $598/month
  • B200 Blackwell (180 GB) — $4,654/month
  • Dedicated GPU, always on
  • Choose your region when ordering
  • All pre-configured apps included
Bare-Metal

Dedicated Servers

On quote
Our sales team prepares a quote for you
  • Dedicated bare-metal NVIDIA GPU servers
  • Full access to the hardware
  • Configuration tailored to your project
  • Set up by the SAFOZi team
  • Custom contract and billing
Contact salesContacter les ventes
11
Select GPUSélect. GPU
22
ConfigureConfigurer
33
DeployDéployer
Choose GPUChoisir GPU
NVIDIA RTX PRO 6000 Blackwell — 96GB GDDR7 — $1.32/hr/h
NVIDIA B200 Blackwell — 180GB HBM3e — $7.50/hr/h
NVIDIA L40S — 48GB GDDR6 — $1.84/hr/h
NVIDIA RTX 4090 — 24GB GDDR6X — $0.78/hr/h

Powered by hosted.ai