Pyyan / GPU & AI Chips / Groq LPU

Groq LPU

Inference only, ~18× GPU speed.

currentGroq ↗Verified 2026-08-13

750

tokens/sec

Specifications

Vendor	Groq

Official links

Groq official site

What matters

No training capability

Best for low-latency serving.

Compare with

All 5 side by side →

Groq LPU vs B300 Blackwell Ultra →Groq LPU vs B200 →Groq LPU vs MI400 →Groq LPU vs H100 →Groq LPU vs TPU v7 Ironwood →

Also in GPU & AI Chips

B300 Blackwell Ultra · NVIDIA B200 · NVIDIA MI400 · AMD H100 · NVIDIA TPU v7 Ironwood · Google