Live decentralized inference network

Decentralized compute for open source AI.

Access elastic GPU clusters globally or deploy your local hardware as a network provider node. Zero upfront commitments, pay-as-you-go billing, and secure isolation by design.

Live GPU Fleet

Active Hardware Clusters

Real-time resource capacity currently running inference operations or idle in regional pools.

NVIDIA H100 SXM5

80GB HBM3
128 Active Nodes
Pricing$2.12/hr
Latency12ms

NVIDIA A100 PCIe

80GB GDDR6
342 Active Nodes
Pricing$1.18/hr
Latency18ms

NVIDIA RTX 4090

24GB GDDR6X
1280 Active Nodes
Pricing$0.42/hr
Latency24ms

NVIDIA RTX 3090

24GB GDDR6
840 Active Nodes
Pricing$0.28/hr
Latency32ms
Token Billing

API Model Endpoints

Launch serverless API endpoints using popular open weight models hosted directly on the network.

ModelDeveloperParamsContextPriceActive Nodes
Llama 3.1 70B InstructMeta70B128K$0.52142
Llama 3.1 8B InstructMeta8B128K$0.08642
Qwen 2.5 Coder 32BAlibaba32B32K$0.24188
Mistral Large 2Mistral123B128K$0.8896
Mistral 7B Instruct v0.3Mistral7B32K$0.06512
Node Operation

Provide compute resources. Receive rewards in real-time.

Turn your local hardware into an active revenue source. Download the Kielo desktop app, sign in, load a model you already have on disk, and go live on the network. Jobs, token usage, and earnings are tracked inside the client — that's how providers contribute capacity to Kielo.

Automatic Routing: No complex port forwarding required. Built-in reverse proxy keeps connections secure.

Instant Settlements: Earn credits per token processed, settled hourly directly to your organization account.

Flexible Scheduling: Toggle provider availability or limit resources allocated during peak local working hours.

Recommended: run the desktop node

Provider dashboards on the web are for overview only. Install Kielo Node to actually host models, accept jobs, and contribute compute to the network.

kielo-node.sh

# 1. Download and initialize the daemon

$ curl -sf https://get.kielo.ai | sh

# 2. Login using Better Auth securely

$ kielo login --api-key=ba_jske...

# 3. Start serving matching clusters

$ kielo node start --allow-gpu=0

● ACTIVEUptime: 24h 12m
GPU: RTX 4090 (24GB)Temp: 64°C
Tokens Processed:1,480,240
Earnings (EST):$12.42
Live Metrics

Network Vital Statistics

Global network throughput, total capacity, and average routing latency metrics.

Total Network Capacity
4.12PFLOPS
+12.4% vs last week
Avg Time to First Token
184ms
99.9% Route Success
Active Compute Nodes
2,590
24 countries connected
Hourly Queries Serviced
14,842
100% network health
Hourly Network Load IndexUpdated 1m ago
24% Load
28% Load
32% Load
45% Load
52% Load
48% Load
62% Load
54% Load
70% Load
85% Load
92% Load
78% Load
64% Load
52% Load
44% Load
38% Load
42% Load
50% Load
68% Load
88% Load
96% Load
84% Load
72% Load
60% Load
24 hours agoCurrent Hour