Distributed inference from a single endpoint
150+
nodes in
40+
countries
A globally distributed edge platform for lightning-fast inference
Easy access to computational power
Live
We make it easier to find computational power by aggregating fragmented GPU capacity and providing a single entry point to access our distributed network of data centers. By bringing inference at the edge, we are able to reduce response latency and support real-time AI features.
Automated backend infrastructure
Coming soon
mkinf democratizes GenAI model deployment at scale. Our platform provides one easy-to-use endpoint to deploy models closer to your users, while scaling the compute resources utilised to match your real time usage variability.
On Demand Availability
H100
80GB - 16vCPU - 120GB RAM
$2.75
/hour
More than 200 cards
A100
80GB - 12vCPU - 120GB RAM
$1.81
/hour
More than 400 cards
A100
40GB - 8vCPU - 80GB RAM
$1.60
/hour
More than 100 cards
L40s
48GB - 8vCPU - 32GB RAM
$1.25
/hour
More than 600 cards
A40
48GB - 8vCPU - 32GB RAM
$1.10
/hour
More than 200 cards
A6000
48GB - 8vCPU - 32GB RAM
$0.75
/hour
More than 100 cards
Features
Quick App Launch
mkinf takes care of all the infrastructure management so you can focus on your AI app development.
Enterprise Grade
Benefit from a network of Tier 3 and 4 Data Centers, guaranteeing secure and fast environment for computational tasks.
Cost Efficiency
Avoid being locked in by long term fixed contracts and pay for exactly what you need.
Scalability
By giving access to our distributed network, we provide flexible compute to companies with variable GPU requirements.
1
Train model
Train your model on our platform or leverage oper source models
2
Select compute
Choose the hardware that best fit your needs and we’ll take care of the rest
3
Deploy
Upload your model and deploy on hundreds of machines from a single endpoint
Get access to instances
We just released our API! Get in touch for an exclusive early access to our GPUs
The energy grid of data centers for compute power
mkinf optimizes computational power resources by creating a supply network of data centers that can leverage their idle hardware.
Our ability to tap into a vast pool of computational power allows for fast responses even during peak times, increases reliability, and ensures continuous service by redirecting tasks if one data center encounters issues.
Thoughts on inference at the edge
Infrastructure optimizations in AI
Managing GPU requirements from Training to Inference
Choosing the right GPU
NVIDIA L40S vs. A100 and H100