back to local model checker

Qwen 3 235B-A22B

MoEToolsThinking

Alibaba · 235B (22B active) · Mixture of Experts

Massive MoE with 22B active — frontier quality

Apache 2.0
Downloads

680.0K

Likes

1.1K

Released

2025-04

Context

128K tokens

Use cases

chatcodereasoning
Total experts128
Active experts8
Active params22.0B

Your hardware

16 GB Reference Desktop

Quant
Q4_K_M
Memory
120.9 GB
Speed
--
Fit
Too heavy

Quantization options

What is quantization?
QuantBitsVRAMDiskQualityStatus
Q2_K275.7 GB71.8 GBlowToo heavy · ~1 t/s
Q3_K_M3105.8 GB100.5 GBmoderateToo heavy · ~1 t/s
Q4_K_M4120.9 GB114.9 GBgoodToo heavy · ~1 t/s
Q5_K_M5151 GB143.6 GBgoodToo heavy · ~1 t/s
Q6_K6181.1 GB172.4 GBexcellentToo heavy · ~1 t/s
Q8_08241.2 GB229.8 GBexcellentToo heavy · ~1 t/s
F1616482 GB459.6 GBlosslessToo heavy · ~1 t/s

About this model

Qwen 3 235B-A22B (Qwen3-235b-a22b) is a 235B sparse mixture-of-experts (MoE) model activating 22B active per token from Alibaba in the Qwen family, designed primarily for chat, code, reasoning workflows. Native reasoning/thinking tokens expand output length per query, so reserve extra KV cache headroom. Supports structured tool calling; use Q4_K_M or higher when strict JSON schema adherence matters. Supports up to 128K context tokens; start at 4K–8K with Q8_0 KV cache on tight GPUs.

Hardware planning baselines start at 131.3 GB minimum memory and 218.9 GB recommended memory. At Q4_K_M, weights plus baseline runtime overhead use roughly 120.9 GB of RAM/VRAM and 114.9 GB of disk space. At 235B, Qwen 3 235B-A22B belongs firmly in the workstation and multi-GPU tier. Running it comfortably requires 218.9 GB of total memory across dual RTX GPUs, high-memory Apple Silicon (Mac Studio Ultra/Max), or MoE hybrid CPU-RAM offloading—mobile devices and standard 16 GB laptops are too constrained for practical speeds.

Deploy Qwen 3 235B-A22B locally via Ollama (`qwen3:235b-a22b`), LM Studio, or Ollama with Flash Attention enabled. Under Apache License 2.0, review official weight cards before commercial redistribution.

Highlights

  • Local Fit Baseline131.3 GB minimum and 218.9 GB recommended RAM/VRAM.
  • Primary QuantQ4_K_M targets 120.9 GB memory and 114.9 GB disk.
  • Context & Capabilities128K max context · Sparse MoE · Tool calling · Reasoning.

System requirements by device

Memory planning targets 131.3 GB minimum RAM/VRAM and 218.9 GB recommended memory. The baseline quant is Q4_K_M, estimated at 120.9 GB memory and 114.9 GB disk. Actual runtime usage shifts with context length, KV cache precision, and OS background overhead.

DeviceNotes
Mac (Apple silicon)Requires a 48 GB–192 GB Mac Studio or MacBook Pro Max/Ultra with enough unified wired budget for 120.9 GB.
Windows / Linux GPUUse dual/multi-GPU CUDA splitting (--tensor-split) or MoE CPU expert offloading (-ot) paired with 32–128 GB DDR5 system RAM.
Phones & Compact Edge BoardsNot practical on phones (131.3 GB min RAM exceeds Android LMK and iOS Jetsam limits). Pick a 2B–4B model on mobile.

Start with a short session and leave memory for the operating system. A file fitting on disk does not guarantee that it will load in memory.

Estimated memory by quantization

These are rough estimates, not tested minimum requirements. Check the actual file and runtime documentation; context, offloading and operating-system overhead can change the memory needed.

Quant TierEstimated memoryDisk SizeQuality
Q2_K75.7 GB71.8 GBlow
Q3_K_M105.8 GB100.5 GBmoderate
Q4_K_M120.9 GB114.9 GBgood
Q5_K_M151.0 GB143.6 GBgood
Q6_K181.1 GB172.4 GBexcellent
Q8_0241.2 GB229.8 GBexcellent
F16482.0 GB459.6 GBlossless

How to run Qwen 3 235B-A22B on your device

Start with a short prompt and the smallest quant that still gives the quality you need. A clean first run is more useful than loading a larger file with no memory headroom.

Windows

Run Qwen 3 235B-A22B on Multi-GPU or High-RAM Desktop

  1. Plan for 131.3 GB minimum combined memory and 218.9 GB recommended memory for Q4_K_M.
  2. Use llama.cpp `-ot ".ffn_.*_exps.=CPU"` or `--tensor-split` across multiple GPUs so attention and KV cache remain in fast VRAM.
  3. Enable Flash Attention (`-fa on`) and Q8_0 KV cache (`-ctk q8_0 -ctv q8_0`) to prevent context buffers from exceeding GPU headroom.
  4. Avoid attempting to run workstation-scale weights on smartphones or 8 GB laptops; use a 2B–8B checkpoint on portable devices.
Device how-to guide
Mac

Run Qwen 3 235B-A22B on High-Memory Apple Silicon Mac

  1. Verify you are on a 48 GB–192 GB Mac Studio or MacBook Pro Max/Ultra capable of holding 120.9 GB in wired unified memory.
  2. Launch via `llama-server` or LM Studio with MLX/Metal enabled and set initial context conservatively to 4K–8K.
  3. Monitor Activity Monitor Memory Pressure to confirm zero SSD swap thrashing during token generation.
  4. Step down to a lighter quantization (e.g., Q3_K_M or IQ4_XS) if total memory utilization exceeds 75%.
Device how-to guide

Qwen 3 235B-A22B vs alternatives

ModelParamsContextMin RAMLicense
Qwen 3 235B-A22B235B (22B active)128K131.3 GBApache 2.0
Qwen 3 VL 235B-A22B235B (22B active)256K131.3 GBApache 2.0
Qwen 3.8 Flash 128B125B (6B active)128K32 GBApache 2.0
Qwen 3.5 122B-A10B122B (10B active)256K68.2 GBApache 2.0
Qwen 3 Next 80B-A3B80B (3B active)256K44.7 GBApache 2.0
Qwen 3 Coder Next 80B-A3B80B (3B active)256K44.7 GBApache 2.0

Use this table to choose the smallest nearby model that still matches the job. A related model with lower RAM can be the better local choice when the current page fits only at the edge of available memory.

Frequently asked questions

What is Qwen 3 235B-A22B?
Qwen 3 235B-A22B is a 235B sparse mixture-of-experts model (22B active active per token) from Alibaba, supporting up to 128K tokens with 131.3 GB minimum RAM and 218.9 GB recommended RAM.
What is the best way to run Qwen3-235b-a22b locally?
Choose the smallest quant that matches the task, test one short prompt, then compare RAM, context, and nearby models before making Qwen 3 235B-A22B your default local model.
How much VRAM or RAM does Qwen 3 235B-A22B require?
At Q4_K_M, Qwen 3 235B-A22B occupies roughly 120.9 GB of active memory and 114.9 GB on disk. Add 2–6 GB of spare headroom for KV cache buffers, runtime scratch space, and OS apps (218.9 GB recommended).
What is the best way to run Qwen 3 235B-A22B locally?
Use `ollama run qwen3:235b-a22b` for a quick CLI/API setup, or load Q4_K_M in LM Studio or llama.cpp with Flash Attention enabled when you want explicit KV cache and layer control.
Can Qwen 3 235B-A22B run on a consumer laptop or smartphone?
No smartphone or 8–16 GB laptop can hold Qwen 3 235B-A22B cleanly in local memory. It targets multi-GPU NVIDIA workstations, high-RAM Mac Studios, or MoE CPU-RAM hybrid rigs with at least 131.3 GB total memory.
How does Qwen 3 235B-A22B compare to Qwen 3 VL 235B-A22B?
Qwen 3 VL 235B-A22B (235B, 131.3 GB min RAM) sits closest in size. Compare active parameters, context ceiling (128K vs 256K), and license terms (Apache 2.0) before switching.