ER

ERNIE 4.5 21B A3B Thinking

ErnieGGUF by bartowskiUpdated 2025-09-12
Download on Hugging Face
Parameters21.8B
Context128K
Layers28
Architectureernie4_5-moe
Smallest quant0.0 GB
Quick answer

ERNIE 4.5 21B A3B Thinking needs about 14.5 GB of memory at Q4_K_M (a 13.5 GB file) with an 8K context window, so it fits on a 16 GB GPU. This repository has 28 quants, from 0.0 GB to 43.7 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.9 GB.

ERNIE 4.5 21B A3B Thinking – a large‑scale MoE model focused on reasoning

ERNIE 4.5 21B A3B Thinking is a text MoE model with 21.8B total parameters and 3B active parameters per token. It supports 128K context length and is optimized for complex reasoning tasks. The model is released under Apache‑2.0 and can be run with llama.cpp or vLLM.

Good at

  • Advanced reasoning
  • Tool‑aware generation
  • Long‑context comprehension

Limits

  • Requires 80GB VRAM for full precision
  • No vision or audio input
  • Limited to text only

Best for

  • Academic research requiring deep reasoning
  • Tool‑integrated chatbot development
  • Long‑form content generation

Who it’s for. Researchers and developers needing a high‑capacity reasoning model. The model is suitable for teams with access to large GPU resources and who can manage MoE inference.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—0.0 GB 1.0 GB
Fits Get file →
IQ2_XXS5.6 GB 6.5 GB
Fits Get file →
IQ2_XS6.3 GB 7.3 GB
Tight Get file →
IQ2_S6.4 GB 7.3 GB
Tight Get file →
IQ2_M7.2 GB 8.1 GB
Too big Get file →
Q2_K8.1 GB 9.1 GB
Too big Get file →
Q2_K_L8.2 GB 9.1 GB
Too big Get file →
IQ3_XXS9.0 GB 10.0 GB
Too big Get file →
IQ3_XS9.4 GB 10.3 GB
Too big Get file →
Q3_K_S9.9 GB 10.8 GB
Too big Get file →
IQ3_M10.3 GB 11.3 GB
Too big Get file →
Q3_K_M10.3 GB 11.3 GB
Too big Get file →
Q3_K_L10.7 GB 11.6 GB
Too big Get file →
Q3_K_XL10.7 GB 11.7 GB
Too big Get file →
IQ4_XS12.0 GB 12.9 GB
Too big Get file →
IQ4_NL12.6 GB 13.6 GB
Too big Get file →
Q4_012.8 GB 13.8 GB
Too big Get file →
Q4_K_S13.0 GB 14.0 GB
Too big Get file →
Q4_K_M13.5 GB 14.5 GB
Too big Get file →
Q4_K_L13.6 GB 14.5 GB
Too big Get file →
Q4_113.9 GB 14.9 GB
Too big Get file →
Q5_K_S15.2 GB 16.2 GB
Too big Get file →
Q5_K_M15.8 GB 16.7 GB
Too big Get file →
Q5_K_L15.8 GB 16.8 GB
Too big Get file →
Q6_K18.1 GB 19.1 GB
Too big Get file →
Q6_K_L18.1 GB 19.1 GB
Too big Get file →
Q8_023.2 GB 24.2 GB
Too big Get file →
BF1643.7 GB 44.6 GB
Too big Get file →

For 8‑GB VRAM, use IQ2_XXS (5.57 GB file, 6.5 GB memory at 8K context). For 16‑GB VRAM, IQ3_XS (9.35 GB file, 10.3 GB memory) or Q3_K_S (9.85 GB file, 10.8 GB memory). For 32‑GB VRAM, Q4_K_M (13.5 GB file, 14.5 GB memory) or Q5_K_S (15.23 GB file, 16.2 GB memory).

What GPU do you need for ERNIE 4.5 21B A3B Thinking?

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: the best quant that fits is the main file (0.0 GB file, about 1.0 GB in use).
  • 6 GB VRAM: the best quant that fits is the main file (0.0 GB file, about 1.0 GB in use).
  • 8 GB VRAM: the best quant that fits is IQ2_S (6.4 GB file, about 7.3 GB in use).
  • 12 GB VRAM: the best quant that fits is Q3_K_XL (10.7 GB file, about 11.7 GB in use).
  • 16 GB VRAM: the best quant that fits is Q4_1 (13.9 GB file, about 14.9 GB in use).
  • 24 GB VRAM: the best quant that fits is Q6_K_L (18.1 GB file, about 19.1 GB in use).
  • 32 GB VRAM: the best quant that fits is Q8_0 (23.2 GB file, about 24.2 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 Thinking quant should you choose?

Q4_K_M is the usual starting point for ERNIE 4.5 21B A3B Thinking: it balances file size and quality well. Here it is a 13.5 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 the first one in the table at 0.0 GB, with the quality trade-offs that come with fewer bits.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureernie4_5-moe
Parameters21.8B (approximate, mixture of experts)
Active parameters3B
Layers28
Trained context131,072 tokens
KV cache (fp16)56 KB per token

Source

Quantized bybartowski
FamilyErnie
Licenseapache-2.0
Downloads (HF)1,558
Last updated2025-09-12

Frequently asked questions about ERNIE 4.5 21B A3B Thinking

How much VRAM does ERNIE 4.5 21B A3B Thinking need?

About 14.5 GB at Q4_K_M with an 8K context window. That is the 13.5 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 Thinking on an 8 GB GPU?

Yes, with IQ2_S: the 6.4 GB file needs about 7.3 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 Thinking GGUF?

It depends on the quant: file is 0.0 GB, IQ2_XXS is 5.6 GB, IQ2_XS is 6.3 GB, IQ2_S is 6.4 GB, IQ2_M is 7.2 GB, Q2_K is 8.1 GB and 22 more.

What context length does ERNIE 4.5 21B A3B Thinking 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 Thinking GGUF?

The files are on Hugging Face in the bartowski/baidu_ERNIE-4.5-21B-A3B-Thinking-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 Thinking 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?

Yes, using the IQ2_XXS quantization which requires 6.5 GB memory at 8K context.

Does this model support image input?

No, the model is text‑only and does not include vision capabilities.

What license applies to this model?

The model is released under the Apache License 2.0, allowing commercial use.

Other GGUFs of ERNIE-4.5-21B-A3B-Thinking

Similar models

Browse all →

More from bartowski

All by bartowski →

More Ernie models

All Ernie →

Other medium models

All medium models →