← All models

Alibaba Qwen · mixture of experts · Apache-2.0

What runs Qwen3.5 397B-A17B

Qwen's largest open MoE of the 3.5 generation, 397B total with 17B active. Runs locally on 256-512 GB unified-memory machines or EPYC servers with large RAM.

Parameters
397B, 17B active
4-bit file
227.3 GB
Q4_K_M
Needs at 8k context
229.1 GB
context memory estimated
Longest context
262,144 tokens

Context memory: 30,720 bytes per token for the full-attention layers only. Hybrid: 15 of 60 layers are full attention (2 x 15 x 2 KV heads x 256 x 2 = 30,720 bytes/token); 45 Gated DeltaNet layers hold a fixed-size state. Card gives no per-token figure. How context memory works.

Measured on this hardware

Machines that hold the 4-bit file with 8,192 tokens of context and have published runs on the same chip or cards, fastest first. Median of single-user runs with plain decoding.

No published run on a machine that fits it.

Estimated, not measured

These fit but nobody has published a run we could match. The figure is memory bandwidth divided by the bytes read per token, scaled by the share of that ceiling measured runs reach. How estimates work.

MachineFitPriceGeneration
Dell Pro Precision with GB300NVIDIA workstationFits~$159,283~320–409 tok/sestimate
Apple Mac Studio (M5 Ultra, 64-core GPU)Apple unified memoryFits with raised limitassumed limitfrom 256 GB~$5,499–$9,499~38–68 tok/sestimate
Apple Mac Studio (M5 Ultra, 80-core GPU)Apple unified memoryFits with raised limitassumed limitfrom 256 GB~$6,799–$10,799~38–68 tok/sestimate
RTX 5090 + EPYC 9355P with 768 GB DDR5 (MoE offload)Hybrid GPU + RAMGPU + RAM offloadassumed limit~$3,763 + unpriced parts~16–23 tok/sestimate, part in RAM

Fits, no speed figure

Too large for

Apple Mac Studio (M5 Max, 32-core GPU), Apple Mac Studio (M5 Max, 40-core GPU), Apple Mac mini (M5 Pro, 16-core GPU), Apple Mac mini (M5 Pro, 20-core GPU), Apple MacBook Pro 14-inch (M5 Max, 32-core GPU), Apple MacBook Pro 14-inch (M5 Max, 40-core GPU), Apple MacBook Pro 16-inch (M5 Max, 32-core GPU), Apple MacBook Pro 16-inch (M5 Max, 40-core GPU), Framework Desktop (Ryzen AI Max+ 395), Framework Desktop (Ryzen AI Max 385), Framework Desktop (Ryzen AI Max+ PRO 495), GMKtec EVO-X2 (Ryzen AI Max+ 395), Beelink GTR9 Pro (Ryzen AI Max+ 395), Minisforum MS-S1 MAX (Ryzen AI Max+ 395), Corsair AI Workstation 300 (Ryzen AI Max+ 395), HP Z2 Mini G1a (Ryzen AI Max+ PRO 395), NVIDIA DGX Spark, Dell Pro Precision with GB10, Microsoft Surface RTX Spark Dev Box, Dell Pro Max Tower T2 with NVIDIA RTX PRO 6000 Blackwell, tiny corp tinybox red flash, 1x RTX 3090 (used), 2x RTX 3090 (used), 4x RTX 3090 (used) on EPYC 7003, open frame, 1x RTX 5090, 2x RTX 5090, 1x RTX PRO 6000 Blackwell, 2x Radeon RX 7900 XTX, 4x Instinct MI50 32GB (used) on EPYC 7003, 2x Intel Arc Pro B60, RTX 3090 (used) + 128 GB DDR5 on AM5 (MoE offload). A smaller quant may fit: try the calculator.

Files

QuantFormatSizeRepo
UD-IQ1_MGGUF99.5 GBunsloth/Qwen3.5-397B-A17B-GGUF
UD-IQ2_XXSGGUF107.0 GBunsloth/Qwen3.5-397B-A17B-GGUF
UD-IQ2_MGGUF114.6 GBunsloth/Qwen3.5-397B-A17B-GGUF
UD-IQ3_XXSGGUF130.7 GBunsloth/Qwen3.5-397B-A17B-GGUF
Q3_K_MGGUF165.2 GBunsloth/Qwen3.5-397B-A17B-GGUF
UD-Q3_K_XLGGUF166.5 GBunsloth/Qwen3.5-397B-A17B-GGUF
4bitMLX208.5 GBmlx-community/Qwen3.5-397B-A17B-4bit
Q4_K_MGGUF227.3 GBunsloth/Qwen3.5-397B-A17B-GGUF
Q5_K_MGGUF273.5 GBunsloth/Qwen3.5-397B-A17B-GGUF
Q6_KGGUF304.2 GBunsloth/Qwen3.5-397B-A17B-GGUF
Q8_0GGUF392.6 GBunsloth/Qwen3.5-397B-A17B-GGUF
8bitMLX393.0 GBmlx-community/Qwen3.5-397B-A17B-8bit

Sources