Alibaba/Mixture of Experts

Qwen 3.6 35B A3B

Name: Qwen 3.6 35B A3B
Author: Alibaba

Qwen 3.6 35B A3B — hybrid linear/full attention MoE (DeltaNet + full attention), multimodal (text+image+video), 256 experts (8+1 active). Prioritizes agentic coding and thinking preservation over Qwen 3.5.

chatcodingreasoningmultilingualvisionmathtool_use

35B

Parameters (3B active)

256K

Context length

Benchmarks

Quantizations

100K

HF downloads

Architecture

MoE

Released

2026-04-15

Layers

KV Heads

Head Dim

256

Family

qwen

Quantization Options

Quant	Bits	VRAM	Quality
IQ3_XXS	3.25	14.7 GB	low
IQ3_XS	3.5	15.8 GB	low
Q3_K_S	3.64	16.4 GB	low
IQ3_M	3.72	16.8 GB	low
Q3_K_M	4	18.0 GB	low
Q3_K_L	4.25	19.1 GB	low
IQ4_XS	4.37	19.6 GB	moderate
Q4_K_S	4.5	20.2 GB	moderate
Q4_K_M	4.89	21.9 GB	good
Q5_K_S	5.57	24.9 GB	good
Q5_K_M	5.7	25.4 GB	good
Q6_K	6.56	29.2 GB	excellent
Q8_0	8.5	37.7 GB	lossless
FP16	16	70.5 GB	lossless

Select your GPU above to see speed estimates and compatibility for each quantization.

Community Ratings

Loading ratings...

Benchmarks (14)

MMBench92.8

AIME92.7

GPQA Diamond86.0

MMLU-PRO85.2

MMMU81.7

LiveCodeBench80.4

IFEval78.9

SWE-bench73.4

MATH59.7

BBH58.3

BigCodeBench32.3

HLE21.4

MUSR19.1

GPQA15.2

Run this model

▸Easiest way to get started·Beginners

DOCS ↗

curl -fsSL https://ollama.com/install.sh | sh

$ollama run qwen3.6:35b-a3b:q4_k_m

Downloads and runs automatically. Add --verbose for speed stats.

▸ SETUP GUIDE

Auto-setup with fitmyllm CLI

Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.

pip install fitmyllmthen run fitmyllmLearn more

Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines

HuggingFace Ollama Library GGUF Downloads Build Hardware

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.