// RUN ON CONSUMER GPU

Run ERNIE-4.5-21B-A3B on Consumer GPU hardware

22 options ranked by estimated speed at Q4 / 8k context.

HardwareEst tok/sFits 8k?Max ctx~Price
NVIDIA GeForce RTX 5090 → ~132.6 yes 62k ~$1,999
NVIDIA GeForce RTX 4090 → ~74.6 yes 34k ~$1,599
NVIDIA GeForce RTX 3090 Ti → ~74.6 yes 34k ~$1,999
NVIDIA GeForce RTX 5080 → ~71 tight 6k ~$999
NVIDIA GeForce RTX 3090 → ~69.2 yes 34k ~$1,499
NVIDIA GeForce RTX 5070 Ti → ~66.3 tight 6k ~$749
AMD Radeon RX 7900 XTX → ~62.9 yes 34k ~$999
NVIDIA GeForce RTX 4080 → ~54.4 tight 6k ~$1,199
AMD Radeon RX 7900 XT → ~52.4 yes 20k ~$899
NVIDIA GeForce RTX 4070 Ti Super → ~49.7 tight 6k ~$799
AMD Radeon RX 9070 XT → ~42.2 tight 6k ~$599
AMD Radeon RX 7800 XT → ~40.9 tight 6k ~$499
Intel Arc Pro B70 → ~37.2 yes 62k —
Intel Arc A770 16GB → ~34.3 tight 6k ~$329
AMD Radeon RX 6800 XT → ~33.5 tight 6k ~$649
AMD Radeon RX 6900 XT → ~33.5 tight 6k ~$999
NVIDIA GeForce RTX 5060 Ti 16GB → ~33.1 tight 6k ~$429
Intel Arc Pro B60 → ~27.9 yes 34k —
NVIDIA GeForce RTX 4060 Ti 16GB → ~21.3 tight 6k ~$499
NVIDIA DGX Spark → ~20.2 yes 389k ~$3,999
AMD Ryzen AI Max+ 395 → ~16.8 yes 389k ~$1,999
Intel Arc Pro B50 → ~13.7 tight 6k —

Estimate at Q4 (memory-bound). Click any option for its full per-hardware estimate and measured numbers.

→ ERNIE-4.5-21B-A3B overview → All hardware

FAQ

What Consumer GPU hardware runs ERNIE-4.5-21B-A3B?▶

On GenAIList, 22 Consumer GPU options can run ERNIE-4.5-21B-A3B at Q4 — the fastest is the NVIDIA GeForce RTX 5090 at ~132.6 tok/s.

How fast is ERNIE-4.5-21B-A3B on Consumer GPU hardware?▶

Estimated generation speed ranges from ~13.7 to ~132.6 tok/s at Q4 (memory-bandwidth bound). Click any option for its full estimate and any measured numbers.