Quantized with llama.cpp usage: /home/artjoms/llama.cpp/build/bin/llama-quantize [--help] [--allow-requan using importance-matrix calibration on a public multilingual + code + math corpus.

Prompt format

<|im_start|>system
{system}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant

Available quants

Filename	Quant	Size	Notes
Qwen3.6-35B-A3B-Q8_0.gguf	`Q8_0`	34.37 GB	Practically lossless. Closest to source with significant size cut.
Qwen3.6-35B-A3B-Q6_K_L.gguf	`Q6_K`	26.79 GB	Q6_K with Q8_0 embed/output tensors. Near-lossless top tier.
Qwen3.6-35B-A3B-Q6_K.gguf	`Q6_K`	26.56 GB	Near-lossless quality. Recommended for highest practical fidelity.
Qwen3.6-35B-A3B-Q5_K_L.gguf	`Q5_K_M`	23.32 GB	Q5_K_M with Q8_0 embed/output. High quality with small overhead.
Qwen3.6-35B-A3B-Q5_K_M.gguf	`Q5_K_M`	23.03 GB	High quality, balanced size. Recommended general-purpose.
Qwen3.6-35B-A3B-Q5_K_S.gguf	`Q5_K_S`	22.33 GB	Slightly smaller than Q5_K_M with similar quality.
Qwen3.6-35B-A3B-Q4_K_L.gguf	`Q4_K_M`	20.06 GB	Q4_K_M with Q8_0 embed/output. Sweet spot of quality and size.
Qwen3.6-35B-A3B-Q4_K_M.gguf	`Q4_K_M`	19.71 GB	Best size/quality tradeoff. Recommended default.
Qwen3.6-35B-A3B-Q4_K_S.gguf	`Q4_K_S`	18.52 GB	Compact with minor quality loss versus Q4_K_M.
Qwen3.6-35B-A3B-IQ4_NL.gguf	`IQ4_NL`	18.42 GB	Slightly larger than IQ4_XS. Online repacking for ARM CPU inference.
Qwen3.6-35B-A3B-IQ4_XS.gguf	`IQ4_XS`	17.44 GB	Most efficient sub-Q4. Smaller than Q4_K_S with comparable quality.

Calibration

Imatrix generated from reapmix (community calibration mix) — ~400K tokens — multilingual + code + math. This is the same class of public calibration data used by other community GGUF publishers; no claim of unique calibration is made for this release.

*_L and *_XL variants override the output tensor and/or token embedding to Q8_0 (versus the base type), at small extra disk for typically improved output stability at low bit-rates.

Download

Single file:

hf download Krasnopjorovs/Qwen3.6-35B-A3B-Imatrix-GGUF --include "Qwen3.6-35B-A3B-Q4_K_M.gguf" --local-dir .

Whole repo:

hf download Krasnopjorovs/Qwen3.6-35B-A3B-Imatrix-GGUF --local-dir ./Qwen3.6-35B-A3B-gguf

Run

./llama-server -m Qwen3.6-35B-A3B-Q4_K_M.gguf -c 32768 -ngl 99 --host 0.0.0.0 --port 8080

Picking a quant

Q8_0 / Q6_K_L — RAM headroom, want ceiling quality
Q5_K_M / Q4_K_L — workstation default, very small quality loss
Q4_K_M — best general size/quality tradeoff, the default choice
IQ4_XS — smallest with serious quality, for tight RAM
IQ3_M / Q3_K_M — when 8GB-class VRAM is the budget
IQ2_M and below — emergency only, quality degrades visibly

Build info

llama.cpp release: usage: /home/artjoms/llama.cpp/build/bin/llama-quantize [--help] [--allow-requan
Host: iron-z-01 (x86_64)
Generated: 2026-06-10T22:32:10

Credits

Original model by Qwen
Calibration: reapmix (community calibration mix)
llama.cpp by ggerganov and contributors