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

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:

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

Sign up for the Empero newsletter at empero.org for releases, evals, and research notes.

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)