Qwen3.8-35B-A3B — GGUF

Developed by Empero

GGUF quantizations of empero-ai/Qwen3.8-35B-A3B-Distill — a distillation of the Qwen3.8 frontier models into the Qwen3.6-35B-A3B Mixture-of-Experts architecture — for llama.cpp, Ollama, LM Studio, Jan, KoboldCpp, and other stock GGUF runtimes.

This card is about choosing a file and running it. The capability writeup, benchmark results, and best practices live on the main model card.

35B total parameters with ~3B active per token — the MoE sparsity means it runs considerably faster than a dense 35B at the same quant, but the whole weight file still has to fit in RAM or VRAM.

[!Note] Qwen3.6-class models are hybrids: 30 Gated DeltaNet layers and 10 full-attention layers, with 256 experts routed 8-per-token. A recent llama.cpp build with Qwen3.6 / Gated DeltaNet MoE support is required — older builds will fail to load the architecture.

Files

File	Quant	Size	Notes
`Qwen3.8-35B-A3B-IQ2_M.gguf`	IQ2_M	12.558 GB	Smallest usable. Fits a 16 GB card.
`Qwen3.8-35B-A3B-Q2_K.gguf`	Q2_K	13.839 GB	2-bit K-quant; widest runtime support at this size.
`Qwen3.8-35B-A3B-IQ3_M.gguf`	IQ3_M	16.340 GB	Strong quality per byte at 3-bit.
`Qwen3.8-35B-A3B-Q3_K_M.gguf`	Q3_K_M	17.664 GB	Conventional 3-bit K-quant.
`Qwen3.8-35B-A3B-IQ4_XS.gguf`	IQ4_XS	19.628 GB	Near Q4_K_M quality, ~2 GB smaller.
`Qwen3.8-35B-A3B-Q4_K_M.gguf`	Q4_K_M	21.713 GB	Recommended. Best quality/size balance for most users.
`Qwen3.8-35B-A3B-Q5_K_M.gguf`	Q5_K_M	25.348 GB	Higher quality, modest size increase.
`Qwen3.8-35B-A3B-Q6_K.gguf`	Q6_K	29.209 GB	Near-lossless.
`Qwen3.8-35B-A3B-Q8_0.gguf`	Q8_0	37.802 GB	Highest-quality quantization.
`Qwen3.8-35B-A3B-BF16.gguf`	BF16	71.067 GB	Full precision reference.
`mmproj-Qwen3.8-35B-A3B-F16.gguf`	F16	0.899 GB	Vision projector. Pair with any text quant above for image input.

The IQ* quants and the 2/3-bit K-quants are calibrated with an importance matrix, which is what keeps them coherent at these bit-widths.

Sizes are exact decimal GB from the uploaded files (1 GB = 1,000,000,000 bytes).

What fits?

Weight-size guidance at modest context. The KV cache is the dominant cost at long context and may require offload regardless of weight quant:

Quant	Guidance
IQ2_M / Q2_K	16 GB VRAM, or 16 GB system RAM. The smallest that stay coherent.
IQ3_M / Q3_K_M	20-24 GB VRAM, or 24 GB system RAM.
IQ4_XS / Q4_K_M	24 GB VRAM for a full GPU load; comfortable on CPU with 32 GB RAM.
Q5_K_M / Q6_K	32 GB VRAM, or 48 GB system RAM.
Q8_0	48 GB VRAM, or 64 GB system RAM.
BF16	80 GB+ VRAM, or 96 GB system RAM. Reference only.

Because only ~3B parameters are active per token, CPU and partial-offload inference are far more practical here than for a dense model of the same file size.

Usage

llama.cpp

llama-cli -m Qwen3.8-35B-A3B-Q4_K_M.gguf \
  --temp 0.6 --top-p 0.95 --top-k 20 \
  -n 16384 -cnv

Use the built-in chat template (-cnv). The model is a reasoning model: every answer opens with a <think> block, so allow a generous -n and strip the <think>...</think> span for end users.

Vision

Pair the projector with any text quant:

llama-mtmd-cli -m Qwen3.8-35B-A3B-Q4_K_M.gguf \
  --mmproj mmproj-Qwen3.8-35B-A3B-F16.gguf \
  --image photo.jpg -p "Describe this image."

The vision tower is inherited unchanged from the Qwen3.6-35B-A3B base — the distillation was text-only and vision behaviour was not evaluated.

Ollama / LM Studio / Jan / KoboldCpp

Download the GGUF of your choice and load it directly; the chat template is embedded in the file. Recommended sampling: temperature=0.6, top_p=0.95, top_k=20.

Verifying downloads

SHA256SUMS in this repo lists checksums for every GGUF:

sha256sum -c SHA256SUMS --ignore-missing

Provenance & licensing

Quantizations of empero-ai/Qwen3.8-35B-A3B-Distill, a distillation of the Qwen3.8 frontier models into Qwen/Qwen3.6-35B-A3B. Weights are Apache-2.0, inherited from the Qwen base, shared as-is.

Stay in the loop

Support / Donate

If this model helped you, consider supporting the project:

BTC: bc1qx6zepu6sfkvshgdmc4ewu6pk6rpadvpgffpp7v
LTC: ltc1qv2mefzps2vtjcpwfx8xxdrpplrcvltswm68r7x

Acknowledgements

Developed and released by Empero
Base model: Qwen3.6-35B-A3B (Alibaba Qwen team)
GGUF quantization: llama.cpp (ggml-org)

empero-ai/Qwen3.8-35B-A3B-Distill-GGUF