All Systems Go
One architecture.
Every environment. Infinite scale.
AI systems built exclusively for inference.
Scroll
Go big. Go small.
Go anywhere.
Not a better GPU.
A new foundation.
A faster GPU is still a GPU. Persimmons' chiplet architecture is designed from first principles to handle growing models and deep context with unmatched performance.
1 – 4 modules1 – 4
5 – 41 modules5 – 41
42 – 99 modules42 – 99
100+ modules100+
Modules
Intelligent Efficiency
Massive compute with minimal footprint
Cognitive Clarity
Prefill to decode at lightning speed
Flexible Growth
Power that flexes along with you