ERNIE 4.5 models
3 ERNIE 4.5 models on Hugging Face, from 361M to 21.9B parameters. At the precision each one is published in, the smallest needs about 0.8 GB of VRAM (ERNIE-4.5-0.3B-PT, cheapest live fit: RTX 5060 Ti) and the largest about 49.1 GB (ERNIE-4.5-21B-A3B-PT, cheapest live fit: A100). The cheapest way to run ERNIE-4.5-0.3B-PT is $0.110/hr.
ERNIE 4.5 models
| Model | Parameters | Published as | VRAM needed | Live GPU fit | Est. $/hr |
|---|---|---|---|---|---|
| ERNIE-4.5-0.3B-PT | 361M | BF16 | 0.8 GB | RTX 5060 Ti | $0.110/hr |
| ERNIE-4.5-21B-A3B-Thinking | 21.8B | BF16 | 48.8 GB | A100 | $1.21/hr |
| ERNIE-4.5-21B-A3B-PT | 21.9B | BF16 | 49.1 GB | A100 | $1.21/hr |
VRAM is for the precision the model is published in, with the same overhead and an 8,192-token context assumed on every page; see the methodology. The fit shown is the lowest-priced single GPU type that holds the model at that precision, or the lowest-priced multi-GPU set (up to 8) when none does.
