Alibaba/Mixture of Experts

AlibabaQwen3-235B-A22B

chat
235.1B
Parameters (22B active)
40K
Context length
14
Benchmarks
17
Quantizations
674K
HF downloads
Architecture
MoE
Released
2025-04-28
Layers
94
KV Heads
4
Head Dim
128
Family
qwen

Quantization Options

QuantBitsVRAMQuality
IQ2_XXS2.3870.4 GBlow
IQ2_M2.9386.6 GBlow
Q2_K3.1693.4 GBlow
IQ3_XXS3.2596.0 GBlow
IQ3_XS3.5103.3 GBlow
Q3_K_S3.64107.5 GBlow
IQ3_M3.76111.0 GBlow
Q3_K_M4118.0 GBlow
Q3_K_L4.3126.9 GBmoderate
IQ4_XS4.46131.6 GBmoderate
Q4_K_S4.67137.7 GBmoderate
Q4_K_M4.89144.2 GBgood
Q5_K_S5.57164.2 GBgood
Q5_K_M5.7168.0 GBgood
Q6_K6.56193.3 GBexcellent
Q8_08.5250.3 GBlossless
FP1616470.7 GBlossless

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

Benchmarks (14)

Arena Elo1367
MATH-50098.0
MATH96.0
HumanEval95.0
IFEval88.0
GPQA Diamond75.3
MMLU-PRO72.0
AIME71.7
AA Math71.7
GPQA66.0
LiveCodeBench52.4
AA Intelligence25.0
AA Coding22.1
HLE10.6

Run this model

Easiest way to get starteddocs →
curl -fsSL https://ollama.com/install.sh | sh
$ollama run qwen3:235.1b-instruct-q4_k_m

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

Setup guide

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

Find the best GPU for Qwen3-235B-A22B

Build Hardware for Qwen3-235B-A22B

Qwen3-235B-A22B235.1B Parameter Mixture of Experts LLM

Model Specifications

Parameters
235.1B (22B active)
Architecture
Mixture of Experts
Context Length
40K tokens
Capabilities
chat
Release Date
2025-04-28
Provider
Alibaba
Family
qwen

VRAM Requirements

QuantizationBPWVRAMQuality
IQ2_XXS2.3870.4 GB65%
IQ2_M2.9386.6 GB75%
Q2_K3.1693.4 GB78%
IQ3_XXS3.2596.0 GB82%
IQ3_XS3.5103.3 GB84%
Q3_K_S3.64107.5 GB85%
IQ3_M3.76111.0 GB86%
Q3_K_M4118.0 GB88%
Q3_K_L4.3126.9 GB90%
IQ4_XS4.46131.6 GB92%
Q4_K_S4.67137.7 GB93%
Q4_K_M4.89144.2 GB94%
Q5_K_S5.57164.2 GB96%
Q5_K_M5.7168.0 GB96%
Q6_K6.56193.3 GB97%
Q8_08.5250.3 GB100%
FP1616470.7 GB100%

Benchmark Scores

HumanEval95.0
MMLU-PRO72.0
MATH96.0
IFEval88.0
GPQA66.0
Arena Elo1367.0
GPQA Diamond75.3
LiveCodeBench52.4
AIME71.7
MATH-50098.0
HLE10.6
AA Intelligence25.0
AA Coding22.1
AA Math71.7

How to Run Qwen3-235B-A22B

Run Qwen3-235B-A22B locally with Ollama (needs 144.2 GB VRAM at Q4_K_M):

ollama run qwen3:235.1b

Compatible GPUs (13)

GPUs that can run Qwen3-235B-A22B at Q4_K_M quantization: