Models

Qwen3.5 35B A3B

Alibaba Qwen · February 2026

Get

Hugging Face

A 36B, 3.5B active chat model from Alibaba Qwen that reads pictures as well as text.

Rating: 3.9
Downloads: 430K
Likes: 1.1K
Parameters: 36B, 3.5B active
Context: 256K
Licence: apache-2.0

Runs well on M4 · 16 GB

The IQ1_M size from bartowski is the one to get: a 9.4 GB download that uses 10 GB of memory, and writes about 22 words a second.

Ratings

3.9 out of 5, across people's blind votes on its answers.

Text: 1394
Coding: 1435

LM Arena ratings of October 28, 2026 (CC BY 4.0): people vote between two answers without knowing which model wrote which. The bar ends at 1488, the best any open model has. The Arena tests the full model; a small size on a Mac answers a little worse.

Qwen3.5 35B A3B runs well from a Mac with an M1 and 16 GB of memory. 56 of 59 Mac configurations run it well.

Compare with

The nearest models in size.

Sizes

From unsloth/Qwen3.5-35B-A3B-GGUF on Hugging Face. A smaller size is a smaller download and a little less accurate.

Size	Download	Memory at 8K	Get
UD-IQ2_XXS	11 GB	11 GBestimated	Open
UD-IQ2_M	11 GB	12 GBestimated	Open
UD-Q2_K_XL	12 GB	12 GBestimated	Open
UD-IQ3_XXS	13 GB	13 GBestimated	Open
UD-IQ3_S	14 GB	14 GBestimated	Open
Q3_K_S	15 GB	15 GBestimated	Open
Q3_K_M	16 GB	17 GBestimated	Open
UD-Q3_K_XL	17 GB	17 GBestimated	Open
UD-IQ4_XS	17 GB	18 GBestimated	Open
UD-IQ4_NL	18 GB	18 GBestimated	Open
UD-Q4_K_L	20 GB	20 GBestimated	Open
Q4_K_S	21 GB	21 GBestimated	Open
MXFP4_MOE	22 GB	22 GBestimated	Open
Q4_K_M	22 GB	22 GBestimated	Open
UD-Q4_K_XL	22 GB	22 GBestimated	Open
Q5_K_S	25 GB	25 GBestimated	Open
Q5_K_M	26 GB	26 GBestimated	Open
UD-Q5_K_XL	26 GB	27 GBestimated	Open
UD-Q6_K_S	29 GB	29 GBestimated	Open
Q6_K	29 GB	29 GBestimated	Open
UD-Q6_K_XL	32 GB	32 GBestimated	Open
Q8_0	37 GB	37 GBestimated	Open
UD-Q8_K_XL	49 GB	49 GBestimated	Open
BF16	69 GB	70 GBestimated	Open

Run It Anywhere

Qwen3.5 35B A3B is an open model, so any app that runs GGUF files can run it. These take the IQ1_M size from bartowski straight from Hugging Face.

LM StudioMac, Windows, LinuxOpen
JanMac, Windows, LinuxOpen
MstyMac, Windows, LinuxOpen
OllamaMac, Windows, Linuxollama run hf.co/bartowski/Qwen_Qwen3.5-35B-A3B-GGUF:IQ1_M
llama.cppMac, Windows, Linuxllama-server -hf bartowski/Qwen_Qwen3.5-35B-A3B-GGUF:IQ1_M
Docker Model RunnerMac, Windows, Linuxdocker model run hf.co/bartowski/Qwen_Qwen3.5-35B-A3B-GGUF:IQ1_M

On a PC Graphics Card

Video memory	Largest size that fits	Download	Memory at 8K
6 GB	None worth running		needs 12 GB
8 GB	None worth running		needs 12 GB
10 GB	None worth running		needs 12 GB
11 GB	None worth running		needs 12 GB
12 GB	UD-IQ2_M	11 GB	12 GB
16 GB	Q3_K_S	15 GB	15 GB
20 GB	UD-IQ4_NL	18 GB	18 GB
24 GB	UD-Q4_K_XL	22 GB	22 GB
32 GB	Q6_K	29 GB	29 GB
48 GB	Q8_0	37 GB	37 GB

The largest size, of those that keep three bits a weight or more, whose memory at an 8K chat is under 92% of the card's; the rest is the desktop's. Estimated from the files, the same way as for a Mac; speed on a PC is not estimated yet. A size that doesn't fit the card still runs, slower, with part of it in ordinary memory.

On every Mac

Mac	Size	Fit	Memory	Writes
M6 · 16 GB	IQ1_M	Runs well	10 GB	36
M6 · 24 GB	Q3_K_S	Runs well	16 GB	27
M6 · 32 GB	Q4_K_S	Runs well	21 GB	22
M5 · 16 GB	IQ1_M	Runs well	10 GB	34
M5 · 24 GB	Q3_K_S	Runs well	16 GB	26
M5 · 32 GB	Q4_K_S	Runs well	21 GB	21
M5 Pro · 24 GB	Q3_K_S	Runs well	16 GB	48
M5 Pro · 48 GB	UD-Q4_K_XL	Runs great	23 GB	37
M5 Pro · 64 GB	Q6_K	Runs great	30 GB	33
M5 Max · 36 GB	UD-Q4_K_XL	Runs well	24 GB	53
M5 Max · 48 GB	UD-Q4_K_XL	Runs great	23 GB	66
M5 Max · 64 GB	Q6_K	Runs great	30 GB	58
M5 Max · 128 GB	UD-Q8_K_XL	Runs great	52 GB	39
M5 Ultra · 96 GB	Q8_0	Runs great	40 GB	96
M5 Ultra · 256 GB	UD-Q8_K_XL	Runs great	52 GB	68
M5 Ultra · 512 GB	UD-Q8_K_XL	Runs great	52 GB	68
M4 · 16 GB	IQ1_M	Runs well	10 GB	22
M4 · 24 GB	Q3_K_S	Runs well	16 GB	18
M4 · 32 GB	Q4_K_S	Runs well	21 GB	15
M4 Pro · 24 GB	Q3_K_S	Runs well	16 GB	30
M4 Pro · 48 GB	UD-Q4_K_XL	Runs great	23 GB	26
M4 Pro · 64 GB	Q6_K	Runs great	30 GB	22
M4 Max · 36 GB	UD-Q4_K_XL	Runs well	24 GB	33
M4 Max · 48 GB	UD-Q4_K_XL	Runs great	23 GB	38
M4 Max · 64 GB	Q6_K	Runs great	30 GB	34
M4 Max · 96 GB	Q8_0	Runs great	40 GB	29
M4 Max · 128 GB	UD-Q8_K_XL	Runs great	52 GB	26
M3 · 8 GB	IQ1_M	Too big	13 GB
M3 · 16 GB	IQ1_M	Runs well	10 GB	20
M3 · 24 GB	Q3_K_S	Runs well	16 GB	16
M3 Pro · 18 GB	UD-IQ2_M	Runs well	12 GB	25
M3 Pro · 36 GB	UD-Q4_K_XL	Runs well	24 GB	17
M3 Max · 36 GB	UD-Q4_K_XL	Runs well	24 GB	29
M3 Max · 48 GB	UD-Q4_K_XL	Runs great	23 GB	35
M3 Max · 64 GB	Q6_K	Runs great	30 GB	31
M3 Max · 96 GB	Q8_0	Runs great	40 GB	25
M3 Max · 128 GB	UD-Q8_K_XL	Runs great	52 GB	22
M3 Ultra · 96 GB	Q8_0	Runs great	40 GB	46
M3 Ultra · 256 GB	UD-Q8_K_XL	Runs great	52 GB	35
M3 Ultra · 512 GB	UD-Q8_K_XL	Runs great	52 GB	35
M2 · 8 GB	IQ1_M	Too big	13 GB
M2 · 16 GB	IQ1_M	Runs well	10 GB	20
M2 · 24 GB	Q3_K_S	Runs well	16 GB	16
M2 Pro · 16 GB	IQ1_M	Runs well	10 GB	32
M2 Pro · 32 GB	Q4_K_S	Runs well	21 GB	23
M2 Max · 32 GB	Q4_K_S	Runs well	21 GB	36
M2 Max · 64 GB	Q6_K	Runs great	30 GB	30
M2 Max · 96 GB	Q8_0	Runs great	40 GB	30
M2 Ultra · 64 GB	Q6_K	Runs great	30 GB	45
M2 Ultra · 128 GB	UD-Q8_K_XL	Runs great	52 GB	35
M2 Ultra · 192 GB	UD-Q8_K_XL	Runs great	52 GB	35
M1 · 8 GB	IQ1_M	Too big	13 GB
M1 · 16 GB	IQ1_M	Runs well	10 GB	13
M1 Pro · 16 GB	IQ1_M	Runs well	10 GB	26
M1 Pro · 32 GB	Q4_K_S	Runs well	21 GB	19
M1 Max · 32 GB	Q4_K_S	Runs well	21 GB	28
M1 Max · 64 GB	Q6_K	Runs great	30 GB	25
M1 Ultra · 64 GB	Q6_K	Runs great	30 GB	33
M1 Ultra · 128 GB	UD-Q8_K_XL	Runs great	52 GB	28

Writes is words a second, estimated from the Mac's memory speed and the model's size; how, and how wrong.

Packaged by

7 repositories on Hugging Face carry this model for llama.cpp. Outmake reads them all as one model.

Information

Maker: Alibaba Qwen
Released: February 2026
Parameters: 36B, 3.5B active
Longest chat: 256K tokens
Reads pictures: Yes
Licence: apache-2.0
Downloads this month: 429,516
Likes: 1,103
Sizes: 24, 11 to 69 GB
Kind: Maker's release
Source: Qwen/Qwen3.5-35B-A3B

Badge for the model card

Shows on Qwen3.5 35B A3B's Hugging Face card which Macs run it well, linked to this page. It updates every morning.

Files, downloads and likes read from Hugging Face on November 5, 2026 and refreshed every morning. Memory and speed are estimates until someone running Outmake measures this model on that Mac.