ER

ERNIE 4.5 21B A3B Base PT

ErnieGGUF by mradermacherUpdated 2025-07-18
Download on Hugging Face
Parameters22.1B
Context128K
Layers28
Architectureernie4_5-moe
Smallest quant4.6 GB
Quick answer

ERNIE 4.5 21B A3B Base PT needs about 14.4 GB of memory at Q4_K_M (a 13.4 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 23 quants, from 4.6 GB (IQ1_S) to 18.1 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 1.9 GB.

ERNIE 4.5 21B A3B Base PT – large MoE language model

ERNIE 4.5 21B A3B Base PT is a 22.1B parameter model with 3B active experts. It supports 131072 token context and is released under Apache‑2.0. The model is available in multiple GGUF quantizations.

Good at

  • Large context text completion
  • MoE efficient inference
  • High‑quality language generation

Limits

  • No vision support
  • Requires 131072 token context
  • Large VRAM for higher quants

Best for

  • Long‑form content creation
  • Technical documentation generation
  • Research prototyping

Who it’s for. Developers and researchers needing a large, text‑only model with extensive context. The model is suitable for environments that can run GGUF files with sufficient VRAM and RAM.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S4.6 GB 5.5 GB
Fits Get file →
IQ1_M5.1 GB 6.0 GB
Fits Get file →
IQ2_XXS5.9 GB 6.8 GB
Fits Get file →
IQ2_XS6.5 GB 7.5 GB
Tight Get file →
IQ2_S6.6 GB 7.6 GB
Tight Get file →
IQ2_M7.3 GB 8.3 GB
Too big Get file →
Q2_K_S7.5 GB 8.5 GB
Too big Get file →
Q2_K8.1 GB 9.1 GB
Too big Get file →
IQ3_XXS8.6 GB 9.5 GB
Too big Get file →
IQ3_XS9.1 GB 10.1 GB
Too big Get file →
Q3_K_S9.6 GB 10.6 GB
Too big Get file →
IQ3_S9.6 GB 10.6 GB
Too big Get file →
IQ3_M9.7 GB 10.7 GB
Too big Get file →
Q3_K_M10.6 GB 11.6 GB
Too big Get file →
Q3_K_L11.5 GB 12.5 GB
Too big Get file →
IQ4_XS11.8 GB 12.8 GB
Too big Get file →
Q4_012.5 GB 13.5 GB
Too big Get file →
Q4_K_S12.6 GB 13.6 GB
Too big Get file →
Q4_K_M13.4 GB 14.4 GB
Too big Get file →
Q4_113.9 GB 14.8 GB
Too big Get file →
Q5_K_S15.2 GB 16.2 GB
Too big Get file →
Q5_K_M15.7 GB 16.7 GB
Too big Get file →
Q6_K18.1 GB 19.1 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (4.56 GB) or IQ1_M (5.06 GB). For 12 GB VRAM, IQ2_S (6.62 GB) or Q2_K_S (7.54 GB) are suitable. For 16 GB VRAM, IQ3_S (9.62 GB) or Q3_K_S (9.61 GB) provide better quality. Memory estimates are approximate.

What GPU do you need for ERNIE 4.5 21B A3B Base 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: the best quant that fits is IQ1_M (5.1 GB file, about 6.0 GB in use).
  • 8 GB VRAM: the best quant that fits is IQ2_S (6.6 GB file, about 7.6 GB in use).
  • 12 GB VRAM: the best quant that fits is Q3_K_M (10.6 GB file, about 11.6 GB in use).
  • 16 GB VRAM: the best quant that fits is Q4_1 (13.9 GB file, about 14.8 GB in use).
  • 24 GB VRAM: the best quant that fits is Q6_K (18.1 GB file, about 19.1 GB in use).

Which ERNIE 4.5 21B A3B Base PT quant should you choose?

Q4_K_M is the usual starting point for ERNIE 4.5 21B A3B Base PT: it balances file size and quality well. Here it is a 13.4 GB file.

If memory is tight, the smallest file is IQ1_S at 4.6 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureernie4_5-moe
Parameters22.1B (approximate, mixture of experts)
Active parameters3B
Layers28
Trained context131,072 tokens
KV cache (fp16)56 KB per token

Source

Quantized bymradermacher
FamilyErnie
Licenseapache-2.0
Downloads (HF)698
Last updated2025-07-18

Frequently asked questions about ERNIE 4.5 21B A3B Base PT

How much VRAM does ERNIE 4.5 21B A3B Base PT need?

About 14.4 GB at Q4_K_M with an 8K context window. That is the 13.4 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 Base PT on an 8 GB GPU?

Yes, with IQ2_S: the 6.6 GB file needs about 7.6 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 Base PT GGUF?

It depends on the quant: IQ1_S is 4.6 GB, IQ1_M is 5.1 GB, IQ2_XXS is 5.9 GB, IQ2_XS is 6.5 GB, IQ2_S is 6.6 GB, IQ2_M is 7.3 GB and 17 more.

What context length does ERNIE 4.5 21B A3B Base 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 Base PT GGUF?

The files are on Hugging Face in the mradermacher/ERNIE-4.5-21B-A3B-Base-PT-i1-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 Base 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 IQ1_S on 8 GB VRAM?

Yes, IQ1_S is 4.56 GB and requires about 5.5 GB at 8k context, fitting within 8 GB VRAM.

Does the model support image inputs?

No, the Base model only supports text completion.

What license applies to this model?

The model is released under the Apache‑2.0 license.

Other GGUFs of ERNIE-4.5-21B-A3B-Base-PT

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Ernie models

All Ernie →

Other medium models

All medium models →