vLLM Inference
Serve Qwen, Llama, and Mistral with an OpenAI-compatible API. Auto token streaming included.
GPUs available
GPU compute on tap.
Africa’s first marketplace.
Rent verified, production-ready GPUs in seconds. Or earn money by sharing yours. Per-minute billing, zero lock-in, local latency.
RTX 4090 / A100
South Africa
Per-minute billing
From R12/hr
No credit card required · Free trial credits included
—Compute hours
—Verified GPUs
—Fleet VRAM
—Instances launched
8Templates
—Nodes registered
Build, launch, and automate without leaving South Africa
MafutaAI's control plane keeps compute local while exposing the tooling and automation your team expects.
Marketplace
Rent GPUs on-demand
Filter by VRAM, architecture, region, and price. Launch dedicated instances or interruptible savings tiers.
OpenTemplates
Launch a template
Pick from CUDA, PyTorch, vLLM, Ollama, ComfyUI, Stable Diffusion, and notebook stacks—curated by MafutaAI to remove setup friction.
OpenDevelopers
Automate with the REST API
Use REST endpoints and curl to script rentals, manage leases, and stream GPU metrics. Full OpenAPI schema included.
OpenTemplate spotlight
Preview a subset of our launch-ready stacks. The full catalogue lives in the templates hub.
PyTorch 2.3 + Lightning
Optimised training image with NCCL, TensorBoard, and multi-GPU awareness baked in.
ComfyUI Studio
Diffusion-first workflow with ControlNet extensions and VRAM telemetry for creative teams.
For providers
Earn ZAR by plugging your GPUs into the MafutaAI marketplace
Connect Linux or Windows hosts (via WSL2) using our node agent. You stay in control of availability windows, pricing, and compliance—tailored specifically for African infrastructure.
Secure tunnel agent
Deploy the MafutaAI container, register your node, and keep workloads isolated via WireGuard.
Set your price
Expose on-demand and interruptible rates in ZAR. Billing, usage, and payouts are handled automatically.
Monitor utilisation
Dashboard metrics cover VRAM, thermals, and earnings so you can optimise your fleet.
Control plane highlights
Everything runs on sovereign infrastructure with telemetry, governance, and automation built in.
Marketplace economics
Hosts set hourly pricing and availability windows. Customers book instances with minute-level billing.
Instance lifecycle
Start, stop, and restart instances through the console or API. Usage and logs stream back in real time.
Telemetry-native
Prometheus metrics feed dashboards and alerts so you always have live insight into VRAM and throughput.
Compliance-ready
Keep data residency in South Africa with audited providers and MFA-secured access to every instance.
Common questions, straight answers
Everything you need to know before getting started.