Ernie-4.5-21B-A3B-Pt needs about 14.3 GB of memory at Q4_K_M (a 13.3 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 27 quants, from 6.5 GB (TQ1_0) to 43.7 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.9 GB.
21B‑parameter ERNIE‑4.5 MoE model for text tasks
ERNIE‑4.5‑21B‑A3B‑PT is a mixture‑of‑experts model with 21 billion total parameters and 3 billion active parameters. It is trained on 131 072‑token context and released under Apache‑2.0.
Good at
- Large‑context text generation
- Fine‑tuned language understanding
- Efficient inference with 4‑bit quantization
Limits
- Requires 8‑bit or higher quant for full accuracy
- No vision support in this variant
- Large VRAM needed for high‑quality quant
Best for
- Chatbot back‑ends with long context
- Research on MoE architectures
- Deployments needing 8‑k token context
Who it’s for. Developers building LLM services that need large context and MoE efficiency. Researchers exploring mixture‑of‑experts models for text generation.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| TQ1_0 | 6.5 GB | 7.5 GB | Tight | Get file → | |
| IQ1_S | 7.1 GB | 8.0 GB | Too big | Get file → | |
| IQ1_M | 7.2 GB | 8.2 GB | Too big | Get file → | |
| IQ2_XXS | 7.8 GB | 8.7 GB | Too big | Get file → | |
| IQ2_M | 8.0 GB | 9.0 GB | Too big | Get file → | |
| Q2_K | 8.2 GB | 9.1 GB | Too big | Get file → | |
| Q2_K_L | 8.2 GB | 9.1 GB | Too big | Get file → | |
| Q2_K_XL | 8.2 GB | 9.2 GB | Too big | Get file → | |
| IQ3_XXS | 9.5 GB | 10.5 GB | Too big | Get file → | |
| Q3_K_S | 9.6 GB | 10.6 GB | Too big | Get file → | |
| Q3_K_M | 10.5 GB | 11.5 GB | Too big | Get file → | |
| Q3_K_XL | 10.6 GB | 11.5 GB | Too big | Get file → | |
| IQ4_XS | 11.8 GB | 12.8 GB | Too big | Get file → | |
| IQ4_NL | 12.5 GB | 13.4 GB | Too big | Get file → | |
| Q4_0 | 12.5 GB | 13.5 GB | Too big | Get file → | |
| Q4_K_S | 12.5 GB | 13.5 GB | Too big | Get file → | |
| Q4_K_M | 13.3 GB | 14.3 GB | Too big | Get file → | |
| Q4_K_XL | 13.4 GB | 14.4 GB | Too big | Get file → | |
| Q4_1 | 13.8 GB | 14.7 GB | Too big | Get file → | |
| Q5_K_S | 15.1 GB | 16.1 GB | Too big | Get file → | |
| Q5_K_M | 15.6 GB | 16.6 GB | Too big | Get file → | |
| Q5_K_XL | 15.7 GB | 16.7 GB | Too big | Get file → | |
| Q6_K | 18.0 GB | 18.9 GB | Too big | Get file → | |
| Q6_K_XL | 19.3 GB | 20.2 GB | Too big | Get file → | |
| Q8_0 | 23.2 GB | 24.2 GB | Too big | Get file → | |
| Q8_K_XL | 26.6 GB | 27.6 GB | Too big | Get file → | |
| BF16 | 43.7 GB | 44.6 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_K_S (13.5 GB file, ~13.5 GB RAM). For 12 GB VRAM, Q4_K_M (13.33 GB file, ~14.3 GB RAM). For 16 GB VRAM, Q5_K_S (15.14 GB file, ~16.1 GB RAM). Estimates are approximate.
What GPU do you need for Ernie-4.5-21B-A3B-Pt?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: the best quant that fits is IQ1_S (7.1 GB file, about 8.0 GB in use).
- 12 GB VRAM: the best quant that fits is Q3_K_XL (10.6 GB file, about 11.5 GB in use).
- 16 GB VRAM: the best quant that fits is Q4_1 (13.8 GB file, about 14.7 GB in use).
- 24 GB VRAM: the best quant that fits is Q6_K_XL (19.3 GB file, about 20.2 GB in use).
- 32 GB VRAM: the best quant that fits is Q8_K_XL (26.6 GB file, about 27.6 GB in use).
- 48 GB VRAM: the best quant that fits is BF16 (43.7 GB file, about 44.6 GB in use).
Which Ernie-4.5-21B-A3B-Pt quant should you choose?
Q4_K_M is the usual starting point for Ernie-4.5-21B-A3B-Pt: it balances file size and quality well. Here it is a 13.3 GB file.
The Q8_0 file is 23.2 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is TQ1_0 at 6.5 GB, with the quality trade-offs that come with fewer bits. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Ernie-4.5-21B-A3B-Pt
How much VRAM does Ernie-4.5-21B-A3B-Pt need?
About 14.3 GB at Q4_K_M with an 8K context window. That is the 13.3 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run Ernie-4.5-21B-A3B-Pt on an 8 GB GPU?
Yes, with IQ1_S: the 7.1 GB file needs about 8.0 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Ernie-4.5-21B-A3B-Pt GGUF?
It depends on the quant: TQ1_0 is 6.5 GB, IQ1_S is 7.1 GB, IQ1_M is 7.2 GB, IQ2_XXS is 7.8 GB, IQ2_M is 8.0 GB, Q2_K is 8.2 GB and 21 more.
What context length does Ernie-4.5-21B-A3B-Pt support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.5 GB at 8K and 1.9 GB at 32K.
Where can I download Ernie-4.5-21B-A3B-Pt GGUF?
The files are on Hugging Face in the unsloth/ERNIE-4.5-21B-A3B-PT-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does Ernie-4.5-21B-A3B-Pt use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run this model on 8 GB VRAM?
Use Q4_K_S quant; it requires about 13.5 GB VRAM, so 8 GB is insufficient.
Does this model support image inputs?
No, this is a text‑only MoE model; vision support is not included.
What license applies?
The model is released under the Apache‑2.0 license.