Medium-size Mixture-of-Experts model from Baidu's new Ernie 4.5 line of foundation models.
To run the smallest Ernie-4.5, you need at least 12 GB of RAM.
Ernie-4.5 models are available in gguf and mlx formats.

ERNIE 4.5, a new family of large-scale models from Baidu.
Ernie 4.5 21B A3B is a medium-size Mixture-of-Experts model, specializing in general-purpose English and Chinese understanding and generation.
Supports a context length of up to 131,072 tokens.
ERNIE-4.5-21B-A3B is a text MoE Post-trained model, with 21B total parameters and 3B activated parameters for each token. The following are the model configuration details:
| Key | Value |
|---|---|
| Modality | Text |
| Training Stage | Posttraining |
| Params(Total / Activated) | 21B / 3B |
| Layers | 28 |
| Heads(Q/KV) | 20 / 4 |
| Text Experts(Total / Activated) | 64 / 6 |
| Vision Experts(Total / Activated) | 64 / 6 |
| Shared Experts | 2 |
| Context Length | 131072 |

ERNIE 4.5 is Apache 2.0 licensed.