QW

Qwen3.8 4b Heretic Trial23

QwenGGUF by hakergeniuszUpdated 2026-10-10
Download on Hugging Face
Parameters4.21B
Context256K
Layers32
Architectureqwen35
Smallest quant2.5 GB
Quick answer

Qwen3.8 4b Heretic Trial23 needs about 4.3 GB of memory at Q4_K_M (a 2.7 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 4 quants, from 2.5 GB (IQ4_XS) to 8.4 GB (BF16). Longer context needs more: at 32K tokens the KV cache alone adds about 4.3 GB.

Qwen3.8 4b Heretic Trial23 – a 4B distilled Qwen model with 262,144‑token context

Qwen3.8 4b Heretic Trial23 is a 4.21B parameter model distilled from Qwen3.8 2.4T A95B. It supports a native 262,144‑token context and is available in several GGUF quantizations. The model is released under the Apache‑2.0 license.

Good at

  • Distilled chain‑of‑thought reasoning
  • Large native context
  • Full fine‑tune of all parameters

Limits

  • No vision support
  • No MTP or speculative decoding
  • Requires recent llama.cpp with Qwen3.5 support

Best for

  • Long‑form reasoning tasks
  • Complex instruction following
  • Large‑context document analysis

Who it’s for. Developers and researchers who need a 4B model with extended context for reasoning. The model is suitable for local deployment on consumer GPUs with sufficient VRAM.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
IQ4_XS2.5 GB 4.1 GB
Fits Get file →
Q4_K_M2.7 GB 4.3 GB
Fits Get file →
BF162.7 GB 4.3 GB
Fits Get file →
BF168.4 GB 10.0 GB
Too big Get file →

For 8 GB VRAM, use the IQ4_XS build (2.53 GB file, ~4.1 GB VRAM at 8 k context). For 12 GB VRAM, the Q4_K_M or BF16 2.71 GB builds are suitable (~4.3 GB VRAM). For 16 GB VRAM, the BF16 8.42 GB build can be used (~10 GB VRAM).

What GPU do you need for Qwen3.8 4b Heretic Trial23?

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 Q4_K_M (2.7 GB file, about 4.3 GB in use).
  • 8 GB VRAM: the best quant that fits is Q4_K_M (2.7 GB file, about 4.3 GB in use).
  • 12 GB VRAM: the best quant that fits is BF16 (8.4 GB file, about 10.0 GB in use).

Which Qwen3.8 4b Heretic Trial23 quant should you choose?

Q4_K_M is the usual starting point for Qwen3.8 4b Heretic Trial23: it balances file size and quality well. Here it is a 2.7 GB file.

If memory is tight, the smallest file is IQ4_XS at 2.5 GB, with the quality trade-offs that come with fewer bits. The usual sweet spot between file size and quality, and the quant most people start with.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen35
Parameters4.21B
Layers32
Trained context262,144 tokens
KV cache (fp16)128 KB per token

Source

Quantized byhakergeniusz
FamilyQwen
Licenseapache-2.0
Downloads (HF)3,174
Last updated2026-10-10

Frequently asked questions about Qwen3.8 4b Heretic Trial23

How much VRAM does Qwen3.8 4b Heretic Trial23 need?

About 4.3 GB at Q4_K_M with an 8K context window. That is the 2.7 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 Qwen3.8 4b Heretic Trial23 on an 8 GB GPU?

Yes, with Q4_K_M: the 2.7 GB file needs about 4.3 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Qwen3.8 4b Heretic Trial23 GGUF?

It depends on the quant: IQ4_XS is 2.5 GB, Q4_K_M is 2.7 GB, BF16 is 2.7 GB, BF16 is 8.4 GB.

What context length does Qwen3.8 4b Heretic Trial23 support?

The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 1.1 GB at 8K and 4.3 GB at 32K.

Where can I download Qwen3.8 4b Heretic Trial23 GGUF?

The files are on Hugging Face in the hakergeniusz/Qwen3.8-4B-Heretic-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 Qwen3.8 4b Heretic Trial23 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, the IQ4_XS quantization requires about 4.1 GB VRAM at 8 k context, which fits in 8 GB.

Does this model support vision inputs?

No, it is a text‑only causal language model.

Can I use speculative decoding with this GGUF?

No, the GGUFs contain no MTP weights and cannot perform speculative decoding.

Other GGUFs of Qwen3.8-4B-Distill

Similar models

Browse all →

More from hakergeniusz

All by hakergeniusz →

More Qwen models

All Qwen →

Other small models

All small models →