AlexAtomic/qwen36-27b-GGUF

Base model: Qwen/Qwen3.6-27B

Qwen3.6 27B, self-quantized to GGUF by Atomic Chat. Built straight from Qwen's original weights with a per-tensor importance matrix. Runs fully offline.

Highlights

First open-weight Qwen3.6 variant, following the Qwen3.5 series, with a focus on stability and real-world utility.
Agentic coding that handles frontend workflows and repository-level reasoning with greater fluency and precision.
Thinking Preservation, a new option to retain reasoning context from historical messages to streamline iterative development.
Base model is multimodal (vision encoder); these GGUF quants cover the text path.
262,144-token native context, extensible up to ~1,010,000 tokens.
Full quant ladder with an importance matrix on every quant over calibration_datav3.

[!NOTE] These GGUFs are self-quantized from the original weights, not a repack. The importance matrix keeps low-bit quants closer to the full-precision model.

[!IMPORTANT] Always pass --jinja so the Qwen3.6 27B chat template is applied. Without it the model can emit malformed turns.

Model Overview

Property	Value
Base model	`Qwen/Qwen3.6-27B`
Total parameters	27B
Layers	64
Context length	262,144 native, extensible up to ~1,010,000
Architecture	Causal LM with vision encoder (Gated DeltaNet + Gated Attention)
This repo	GGUF quants (imatrix), text path

Scores are Qwen's published results for the base Qwen/Qwen3.6-27B. Quantization preserves the large majority of this; Q4_K_M and up sit within a point or two of full precision.

Choosing a quant

Quant	Size	Notes
`Q2_K`	10.7 GB	Smallest. Minimal RAM, clear quality drop.
`IQ3_M`	12.6 GB	Beats Q3 at similar size thanks to imatrix. Best low-RAM pick.
`Q3_K_M`	13.3 GB	Low quality but usable.
`Q3_K_L`	14.3 GB	A step above Q3_K_M.
`IQ4_XS`	15.1 GB	Excellent quality for size. Recommended low-bit.
`Q4_K_S`	15.6 GB	Compact Q4, fast.
`Q4_K_M`	16.5 GB	Recommended default. Best balance of size, speed and quality.
`UD-Q4_K_XL`	17.5 GB	Dynamic. Embeddings and output kept at Q8_0 for higher quality at a Q4 footprint.
`Q5_K_S`	18.7 GB	Higher quality.
`Q5_K_M`	19.2 GB	Higher quality, low loss.
`Q6_K`	22.1 GB	Near lossless.
`Q8_0`	28.6 GB	Effectively lossless, reference quality.

[!TIP] Pick the largest file that fits your (V)RAM with room for context. Q4_K_M or UD-Q4_K_XL is the sweet spot for most setups; Q6_K or Q8_0 for maximum fidelity.

Get started

Run Qwen3.6 27B locally with:

Atomic Chat: the easiest path. Open the app, search AlexAtomic/qwen36-27b-GGUF, pick a quant, hit Use this model.
llama.cpp: llama-server -hf AlexAtomic/qwen36-27b-GGUF:Q4_K_M --jinja -c 8192
Ollama: ollama run hf.co/AlexAtomic/qwen36-27b-GGUF:Q4_K_M
LM Studio / Jan: search the repo id, download any quant.

Best practices

Parameter	Value
temperature	0.7
top_p	0.8
top_k	20
min_p	0.0
presence_penalty	1.5
repetition_penalty	1.0

Qwen's recommended Instruct (non-thinking) settings. Thinking mode for general tasks: temperature=1.0, top_p=0.95, top_k=20, min_p=0.0, presence_penalty=0.0, repetition_penalty=1.0.

Run in llama.cpp

git clone https://github.com/ggerganov/llama.cpp
cmake llama.cpp -B llama.cpp/build -DBUILD_SHARED_LIBS=OFF -DGGML_CUDA=ON
cmake --build llama.cpp/build --config Release -j --target llama-cli llama-server

./llama.cpp/build/bin/llama-server \
    -hf AlexAtomic/qwen36-27b-GGUF:UD-Q4_K_XL \
    --jinja -ngl 99 -c 8192 -fa on

How these were made

Download Qwen/Qwen3.6-27B (original weights).
Convert to f16 GGUF with llama.cpp.
Build an importance matrix over calibration_datav3 (100 chunks).
Quantize the full ladder with --imatrix.
UD-Q4_K_XL additionally pins the token-embedding and output tensors to Q8_0.

License

Released by Qwen under the Apache 2.0 license. Quantized by Atomic Chat.