How much memory do I need to run Qwen3.5-397B-A17B (Q4_K_M)?

The file size is 244 GB (measured). With a 8,192-token context, the estimated memory including the KV cache is 246 GB. The formula is published at https://localllm.az-arte.com/en/guide/memory-calculation.

Does Qwen3.5-397B-A17B run in 8 GB, 12 GB, 16 GB or 24 GB of VRAM?

8 GB: no quantization fits; 12 GB: no quantization fits; 16 GB: no quantization fits; 24 GB: no quantization fits (estimated at 8,192 tokens with an f16 KV cache, files from the most-downloaded source).

What is the command to run Qwen3.5-397B-A17B with Ollama?

Run "ollama run hf.co/unsloth/Qwen3.5-397B-A17B-GGUF:Q4_K_M" to use the GGUF on Hugging Face directly. To use another quantization, replace the name at the end.

Qwen

Qwen3.5-397B-A17B

Parameters: 396.3B
Max context length: 262,144
Architecture: qwen35moe
License: Commercial OK

Base model: Qwen/Qwen3.5-397B-A17B
License: apache-2.0
llama.cpp: b7990+

Memory needed to run Q4_K_M at 8K tokens

246 GB Estimated

File size 244 GB Measured

VRAM it fits in (rough guide)

8 GB does not fit
12 GB does not fit
16 GB does not fit
24 GB does not fit
32 GB does not fit
48 GB does not fit

The smallest, UD-IQ1_M, needs an estimated 108 GB at 8K.

Macs cap how much memory the GPU can use, so leave headroom in unified memory.

Data as of Oct 11, 2026

About Qwen3.5-397B-A17B

Qwen3.5-397B-A17B is a GGUF model based on Qwen/Qwen3.5-397B-A17B on Hugging Face. GGUF files are published by unsloth/Qwen3.5-397B-A17B-GGUF. It has about 396.3B parameters, the qwen35moe architecture, a maximum context length of 262,144 tokens (values written in the GGUF).

There are 29 quantizations, with file sizes from 107 GB (UD-IQ1_M) to 793 GB (BF16) (actual sizes on Hugging Face).

Running Q4_K_M (244 GB) with an 8,192-token context needs an estimated 246 GB of memory, and an estimated 246 GB at 32,768 tokens. The smallest, UD-IQ1_M (107 GB), needs an estimated 108 GB at 8,192 tokens.

The original model is licensed under Apache License 2.0, which allows commercial use (keep the license text and copyright notice when redistributing). Check the original license for the terms of use. An mmproj file for image input is also available.

File size and estimated memory per quantization

Up to 8 GB
Up to 16 GB
Up to 24 GB
Up to 48 GB
Up to 128 GB
More

File sizes are the actual values on Hugging Face. Memory is estimated for each context length (KV cache f16). The formula is in How memory is calculated.

All memory figures are Estimated (KV cache f16). It should run on a setup with at least this much VRAM (or Mac unified memory). Click a value to see the devices and models that fit that amount (Find by GPU / Mac).

Using image input: This model supports image input. To use images, an mmproj file is loaded as well.

About the "quality guide" (how far a quantization drifts from the original)

For each quantization type, how far the output drifts from the original model, grouped into six levels from values llama.cpp measured on Llama-3-8B.

The value is mean KLD (the difference between the original and quantized model in next-token probabilities); smaller means closer to the original. Quantizations made with an imatrix (importance data) come out slightly smaller. Levels use the value without an imatrix (or with one, for types that only have that).

These values were not measured on this model. Larger models and MoE models tend to drift less with the same type. Distributor-specific mixed types (such as UD-Q4_K_XL) cannot be matched by type name alone, so they show "-".

Quantization	Quality guide	KLD (no imatrix)	KLD (imatrix)	Size (Llama-3-8B)
Q8_0	Nearly identical	0.0014	-	7.96 GiB
Q6_K	Very small	0.0055	-	6.14 GiB
Q5_K_M	Very small	0.0108	-	5.33 GiB
Q5_K_S	Small	0.0166	-	5.21 GiB
Q5_1	Small	0.0180	-	5.65 GiB
Q5_0	Small	0.0222	-	5.21 GiB
Q4_K_M	Small	0.0313	0.0282	4.58 GiB
Q4_K_S	Medium	0.0431	0.0320	4.37 GiB
IQ4_NL	Small	-	0.0357	4.35 GiB
IQ4_XS	Small	-	0.0363	4.14 GiB
Q4_1	Medium	0.0717	-	4.78 GiB
Q4_0	Medium	0.0719	-	4.34 GiB
Q3_K_L	Medium	0.0872	0.0731	4.03 GiB
Q3_K_M	Medium	0.1019	0.0844	3.74 GiB
IQ3_M	Medium	-	0.1025	3.53 GiB
IQ3_S	Medium	-	0.1113	3.42 GiB
IQ3_XS	Large	-	0.1387	3.28 GiB
IQ3_XXS	Large	-	0.1836	3.05 GiB
Q3_K_S	Large	0.2282	0.1998	3.41 GiB
IQ2_M	Large	-	0.3260	2.74 GiB
Q2_K	Very large	0.4451	0.3322	2.96 GiB
Q2_K_S	Very large	-	0.4034	2.96 GiB
IQ2_S	Very large	-	0.4392	2.56 GiB
IQ2_XS	Very large	-	0.5463	2.43 GiB
IQ2_XXS	Very large	-	0.8120	2.24 GiB
IQ1_M	Very large	-	1.3931	2.01 GiB
IQ1_S	Very large	-	2.2546	1.88 GiB

Source: llama.cpp (MIT), measured on Llama-3-8B against FP16, Wikitext-2, revision f364eb6f. Original table · Reading and choosing quantizations

GGUF source

unsloth/Qwen3.5-397B-A17B-GGUF

Converted and published by a third party About this label

Last checked: Oct 11, 2026

Architecture: qwen35moe; llama.cpp b7990 and later
Parameters: 396.3B
Max context length: 262,144
License: apache-2.0 Commercial OK

File size and memory per quantization for Qwen3.5-397B-A17B (unsloth/Qwen3.5-397B-A17B-GGUF)
Quantization	File size Measured	Quality guide	4,096	8,192	16,384	32,768	65,536	131,072	262,144 max
UD-IQ1_M (4 parts)	107 GB	-	108 GB	108 GB	109 GB	109 GB	110 GB	112 GB	116 GB
UD-IQ2_XXS (4 parts)	115 GB	-	116 GB	116 GB	117 GB	117 GB	118 GB	120 GB	124 GB
UD-IQ2_M (4 parts)	123 GB	-	124 GB	125 GB	125 GB	125 GB	126 GB	128 GB	132 GB
UD-IQ3_XXS (4 parts)	140 GB	-	142 GB	142 GB	142 GB	143 GB	144 GB	146 GB	150 GB
UD-IQ3_S (4 parts)	146 GB	-	148 GB	148 GB	148 GB	149 GB	150 GB	152 GB	156 GB
UD-Q3_K_S (5 parts)	164 GB	-	166 GB	166 GB	166 GB	167 GB	168 GB	170 GB	174 GB
Q3_K_S (5 parts)	164 GB	Large	166 GB	166 GB	166 GB	167 GB	168 GB	170 GB	174 GB
UD-Q3_K_M (5 parts)	177 GB	-	179 GB	179 GB	179 GB	180 GB	181 GB	183 GB	187 GB
Q3_K_M (5 parts)	177 GB	Medium	179 GB	179 GB	179 GB	180 GB	181 GB	183 GB	187 GB
UD-Q3_K_XL (5 parts)	179 GB	-	180 GB	180 GB	180 GB	181 GB	182 GB	184 GB	188 GB
UD-IQ4_XS (5 parts)	190 GB	-	191 GB	191 GB	191 GB	192 GB	193 GB	195 GB	199 GB
UD-IQ4_NL (5 parts)	194 GB	-	195 GB	195 GB	196 GB	196 GB	197 GB	199 GB	203 GB
Q4_K_S (6 parts)	228 GB	Medium	229 GB	229 GB	230 GB	230 GB	231 GB	233 GB	237 GB
UD-Q4_K_S (6 parts)	228 GB	-	229 GB	229 GB	230 GB	230 GB	231 GB	233 GB	237 GB
MXFP4_MOE (6 parts)	237 GB	-	239 GB	239 GB	239 GB	240 GB	241 GB	243 GB	247 GB
Q4_K_M (6 parts)	244 GB	Small	245 GB	246 GB	246 GB	246 GB	247 GB	249 GB	253 GB
UD-Q4_K_M (6 parts)	244 GB	-	245 GB	246 GB	246 GB	246 GB	247 GB	249 GB	253 GB
UD-Q4_K_XL (6 parts)	245 GB	-	247 GB	247 GB	247 GB	248 GB	249 GB	251 GB	255 GB
UD-Q5_K_S (7 parts)	277 GB	-	278 GB	278 GB	278 GB	279 GB	280 GB	282 GB	286 GB
Q5_K_S (7 parts)	277 GB	Small	278 GB	278 GB	278 GB	279 GB	280 GB	282 GB	286 GB
UD-Q5_K_M (8 parts)	294 GB	-	295 GB	295 GB	295 GB	296 GB	297 GB	299 GB	303 GB
Q5_K_M (8 parts)	294 GB	Very small	295 GB	295 GB	295 GB	296 GB	297 GB	299 GB	303 GB
UD-Q5_K_XL (8 parts)	295 GB	-	296 GB	296 GB	297 GB	297 GB	298 GB	300 GB	304 GB
Q6_K (8 parts)	327 GB	Very small	328 GB	328 GB	328 GB	329 GB	330 GB	332 GB	336 GB
UD-Q6_K (8 parts)	327 GB	-	328 GB	328 GB	328 GB	329 GB	330 GB	332 GB	336 GB
UD-Q6_K_XL (9 parts)	362 GB	-	364 GB	364 GB	364 GB	365 GB	366 GB	368 GB	372 GB
Q8_0 (10 parts)	422 GB	Nearly identical	423 GB	423 GB	423 GB	424 GB	425 GB	427 GB	431 GB
UD-Q8_K_XL (10 parts)	428 GB	-	429 GB	429 GB	429 GB	430 GB	431 GB	433 GB	437 GB
BF16 (17 parts)	793 GB	Original precision	794 GB	794 GB	795 GB	795 GB	796 GB	798 GB	802 GB

The mmproj file for image input (F16) is 0.92 GB Measured . It is not included in the memory above, so add it if you use images.

Ollama  ollama run hf.co/unsloth/Qwen3.5-397B-A17B-GGUF:Q4_K_M

llama.cpp  llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:Q4_K_M

Third-party evaluations

Values from other organizations for the original model before quantization. Not this site's own rating. Values are for the original model before quantization. GGUF quantization (especially small quantizations) can lower performance.

Human preference (LMArena)

1,441 1,438–1,445

qwen3.5-397b-a17b · 97,678 votes · Values as of Nov 9, 2026

Source: LMArena (CC BY 4.0). Original leaderboard · See in ranking

Capabilities (Epoch AI)

147.9 145.8–150.2

Qwen3.5 397B-A17B · Values as of Feb 13, 2026

Source: Epoch Capabilities Index (Epoch AI) (CC BY 4.0). Original leaderboard · See in ranking

Hard science (GPQA)

85.9% 81.0%–90.7%

qwen3.5-397b-a17b · Values as of Aug 7, 2026

Source: GPQA Diamond (Epoch AI) (CC BY 4.0). Original leaderboard · See in ranking

License and commercial use

Apache License 2.0 Commercial OK

Allows commercial use, modification and redistribution. When redistributing, keep the license text and copyright notice, and note changes in modified files.

Based on the license on the original model's Hugging Face page. Original license text (checked Oct 18, 2026)

The summary states facts this site read from the original text and is not legal advice. Always check the original license and its latest terms before use. All licenses

How to run

Mainline llama.cpp can load this model's format (qwen35moe) from b7990 (Feb 10, 2026) onward. Ollama and LM Studio bundle llama.cpp internally, so an older app fails to load it with errors such as "unknown model architecture". If that happens, update the app.

Example commands using Q4_K_M. To use another quantization, replace the name with one from the table above. How to run local LLMs

Ollama

Run it by pointing directly at the GGUF on Hugging Face.

ollama run hf.co/unsloth/Qwen3.5-397B-A17B-GGUF:Q4_K_M

llama.cpp

Pass the repository and quantization name to llama-server (or llama-cli) with -hf.

llama-server -hf unsloth/Qwen3.5-397B-A17B-GGUF:Q4_K_M

LM Studio

Paste unsloth/Qwen3.5-397B-A17B-GGUF into the search box, download it, and choose the quantization from the file list.

Frequently asked questions

How much memory do I need to run Qwen3.5-397B-A17B (Q4_K_M)?: The file size is 244 GB (measured). With a 8,192-token context, the estimated memory including the KV cache is 246 GB. The formula is published at https://localllm.az-arte.com/en/guide/memory-calculation.
Can Qwen3.5-397B-A17B be used commercially?: The original model's license is Apache License 2.0 (Commercial use allowed).Allows commercial use, modification and redistribution. When redistributing, keep the license text and copyright notice, and note changes in modified files. Check the original license text before deciding.
Does Qwen3.5-397B-A17B run in 8 GB, 12 GB, 16 GB or 24 GB of VRAM?: 8 GB: no quantization fits; 12 GB: no quantization fits; 16 GB: no quantization fits; 24 GB: no quantization fits (estimated at 8,192 tokens with an f16 KV cache, files from the most-downloaded source).
What is the command to run Qwen3.5-397B-A17B with Ollama?: Run "ollama run hf.co/unsloth/Qwen3.5-397B-A17B-GGUF:Q4_K_M" to use the GGUF on Hugging Face directly. To use another quantization, replace the name at the end.