VE

Velvet Eclipse v0.1 4x12B MoE

VelvetGGUF by mradermacherUpdated 2025-04-18
Download on Hugging Face
Parameters38.7B
Context1000K
Layers40
Architecturellama
Smallest quant8.3 GB
Quick answer

Velvet Eclipse v0.1 4x12B MoE needs about 25.3 GB of memory at Q4_K_M (a 23.5 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 23 quants, from 8.3 GB (IQ1_S) to 31.7 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 5.4 GB.

Velvet Eclipse v0.1 4x12B MoE is a 38.7B-parameter Velvet model in GGUF format, quantized by mradermacher. Its trained context length is 1000K tokens. The repository has 23 quant files, from 8.3 GB (IQ1_S) to 31.7 GB (Q6_K).

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ1_S8.3 GB 10.2 GB
Too big Get file →
IQ1_M9.1 GB 11.0 GB
Too big Get file →
IQ2_XXS10.5 GB 12.3 GB
Too big Get file →
IQ2_XS11.6 GB 13.4 GB
Too big Get file →
IQ2_S11.9 GB 13.8 GB
Too big Get file →
IQ2_M13.0 GB 14.9 GB
Too big Get file →
Q2_K_S13.4 GB 15.3 GB
Too big Get file →
Q2_K14.4 GB 16.2 GB
Too big Get file →
IQ3_XXS15.1 GB 16.9 GB
Too big Get file →
IQ3_XS16.0 GB 17.9 GB
Too big Get file →
Q3_K_S16.9 GB 18.7 GB
Too big Get file →
IQ3_S16.9 GB 18.8 GB
Too big Get file →
IQ3_M17.2 GB 19.1 GB
Too big Get file →
Q3_K_M18.7 GB 20.5 GB
Too big Get file →
Q3_K_L20.2 GB 22.0 GB
Too big Get file →
IQ4_XS20.8 GB 22.6 GB
Too big Get file →
Q4_022.0 GB 23.9 GB
Too big Get file →
Q4_K_S22.1 GB 24.0 GB
Too big Get file →
Q4_K_M23.5 GB 25.3 GB
Too big Get file →
Q4_124.3 GB 26.2 GB
Too big Get file →
Q5_K_S26.7 GB 28.5 GB
Too big Get file →
Q5_K_M27.5 GB 29.3 GB
Too big Get file →
Q6_K31.7 GB 33.6 GB
Too big Get file →

What GPU do you need for Velvet Eclipse v0.1 4x12B MoE?

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: no quant fits at an 8K context.
  • 12 GB VRAM: the best quant that fits is IQ1_M (9.1 GB file, about 11.0 GB in use).
  • 16 GB VRAM: the best quant that fits is Q2_K_S (13.4 GB file, about 15.3 GB in use).
  • 24 GB VRAM: the best quant that fits is Q4_K_S (22.1 GB file, about 24.0 GB in use).
  • 32 GB VRAM: the best quant that fits is Q5_K_M (27.5 GB file, about 29.3 GB in use).
  • 48 GB VRAM: the best quant that fits is Q6_K (31.7 GB file, about 33.6 GB in use).

Which Velvet Eclipse v0.1 4x12B MoE quant should you choose?

Q4_K_M is the usual starting point for Velvet Eclipse v0.1 4x12B MoE: it balances file size and quality well. Here it is a 23.5 GB file.

If memory is tight, the smallest file is IQ1_S at 8.3 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

Architecturellama
Parameters38.7B (approximate, mixture of experts)
Layers40
Trained context1,024,000 tokens
KV cache (fp16)160 KB per token

Source

Quantized bymradermacher
FamilyVelvet
Licenseapache-2.0
Downloads (HF)179
Last updated2025-04-18

Frequently asked questions about Velvet Eclipse v0.1 4x12B MoE

How much VRAM does Velvet Eclipse v0.1 4x12B MoE need?

About 25.3 GB at Q4_K_M with an 8K context window. That is the 23.5 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 Velvet Eclipse v0.1 4x12B MoE on an 8 GB GPU?

Not at an 8K context. The smallest file needs about 10.2 GB, so you would need a larger GPU or CPU offloading.

What is the file size of Velvet Eclipse v0.1 4x12B MoE GGUF?

It depends on the quant: IQ1_S is 8.3 GB, IQ1_M is 9.1 GB, IQ2_XXS is 10.5 GB, IQ2_XS is 11.6 GB, IQ2_S is 11.9 GB, IQ2_M is 13.0 GB and 17 more.

What context length does Velvet Eclipse v0.1 4x12B MoE support?

The GGUF header lists a trained context length of 1,024,000 tokens (1000K). The KV cache grows with the context you use: about 1.3 GB at 8K and 5.4 GB at 32K.

Where can I download Velvet Eclipse v0.1 4x12B MoE GGUF?

The files are on Hugging Face in the mradermacher/Velvet-Eclipse-v0.1-4x12B-MoE-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 Velvet Eclipse v0.1 4x12B MoE 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.

Other GGUFs of Velvet-Eclipse-v0.1-4x12B-MoE

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Velvet models

All Velvet →

Other large models

All large models →