The lab
A tour of the Nova platform.
Interactive demos and dashboard patterns we use across every engagement.
Nova · Humanoid Core
A living interface for every model.
Move your cursor — Nova follows
Live network
42M inferences · 6 regions · sub-200ms
Every request routed to the optimal region, model, and cache tier — in real time.
Nova Agent
gpt-nova-4 · streaming
How would you route a spike in inference traffic?
I'd shift 40% of the load to the edge region with the lowest p99, hold speculative decoding for cold prompts, and batch anything under 8k tokens. Want me to draft the runbook?
Overview
Model performance
184ms
Latency p95
12.4k/s
Throughput
62%
GPU util
Requests · last 24h
1,284,921