llmrun

Qwen 3 Models — Hardware Requirements

79 Qwen 3 models from Alibaba and the community, from the smallest that runs in 0.8 GB of VRAM at Q4 up to 480.2B parameters. Every row links to full quantization tables and GPU compatibility.

All Qwen 3 Models by Size

ModelParamsContext
Qwen3 4B DFlash B16537M41K
Qwen3 Embedding 0.6B596M33K
Qwen3 0.6B Base596M33K
Qwen3 0.6B752M41K
Qwen3Guard Gen 0.6B752M33K
Qwen3 8B Speculator.eagle31.0B—
Qwen3 1.7B Base1.7B33K
Qwen3 1.7B2.0B41K
Qwen3 VL 2B Instruct2.1B262K
Qwen3 VL 2B Thinking2.1B262K
Qwen3 4B Z Image Engineer V44B—
Qwen3 4B Instruct 25074.0B262K
Qwen3 4B4.0B41K
Qwen3 4B Thinking 25074.0B262K
Qwen3 4B Base4.0B33K
Typhoon2.5 Qwen3 4B4.0B262K
Qwen3 4B Base Dapo V44.0B33K
Qwen3Guard Gen 4B4.4B33K
Qwen3 VL 4B Instruct4.4B262K
Qwen3 VL 4B Thinking4.4B262K
Qwen3 8B8.2B41K
Qwen3 8B Base8.2B33K
Qwen3Guard Gen 8B8.2B33K
Qwen3 VL 8B Instruct8.8B262K
Qwen3 VL 8B Thinking8.8B262K
Qwen3 14B14.8B41K
Qwen3 14B Base14.8B33K
Qwen3 Coder 30B A3B Instruct30.5B262K
Qwen3 30B A3B Instruct 250730.5B262K
Qwen3 30B A3B30.5B41K
Qwen3 30B A3B Thinking 250730.5B262K
Typhoon2.5 Qwen3 30B A3b30.5B262K
Qwen3 30B A3B Base30.5B33K
Qwen3 VL 30B A3B Instruct31.1B262K
Qwen3 Omni 30B A3B Thinking31.7B—
Qwen3 32B32.8B41K
Qwen3 VL 32B Instruct33.4B262K
Qwen3 Omni 30B A3B Instruct35.3B—
Qwen3 Coder Next79.7B262K
Qwen3 Next 80B A3B Instruct81.3B262K
Qwen3 Next 80B A3B Thinking81.3B262K
Qwen3 235B A22B Instruct 2507235.1B262K
Qwen3 235B A22B235.1B41K
Qwen3 235B A22B Thinking 2507235.1B262K
Qwen3 VL 235B A22B Instruct235.7B262K
Qwen3 VL 235B A22B Thinking235.7B262K
Qwen3 Coder 480B A35B Instruct480.2B262K
Community variants and fine-tunes(28)
ModelParamsContext
Qwen3 4B Domino B16588M41K
Qwen3 0.6B Heretic Abliterated Uncensored596M41K
Distil Qwen3 0.6B Text2sql596M41K
Qwen3 14B PARO1.6B41K
Qwen3 1.7B Abliterated1.7B—
Qwen3 4B Gemini 3.1 Pro Reasoning Distilled4B262K
Qwen3 Code Reasoning 4B4B262K
Qwen3 Sex4.0B262K
Qwen3 4B Heretic4.0B41K
Qwen3 4B Instruct 2507 Heretic4.0B262K
Huihui Qwen3 4B Abliterated v24.0B41K
Parable Qwen3 4B Claude Fable 54.0B41K
Qwen3 4B Abliterated4.0B—
Qwen3 4B Hindi Instruct v24.0B262K
Qwen3 4B Mamba3 Hybrid Stage2b Kd Bias04.7B41K
Huihui Qwen3 8B Abliterated v28.2B41K
Qwen3 8B 64k Context 2X Josiefied Uncensored8.2B66K
Qwen3 8B Heretic8.2B41K
Josiefied Qwen3 8B Abliterated V18.2B41K
Qwen3 8B Abliterated8.2B—
Qwen3 VL 8B Heretic 1.3.08.8B262K
Huihui Qwen3 Coder 30B A3B Instruct Abliterated30.5B262K
Huihui Qwen3 30B A3B Instruct 2507 Abliterated30.5B262K
Qwen3 30B A3B Abliterated30.5B41K
Qwen3 30B A3B.w8a830.6B41K
Qwen3 42B A3B 2507 Thinking Abliterated Uncensored TOTAL RECALL v2 Medium MASTER CODER42.4B262K
Huihui Qwen3 Coder Next Abliterated79.7B262K
Qwen3 Nemotron 235B A22B GenRM 2603235.1B262K

How Qwen 3 Compares — llmrun Composite Score

Qwen3 235B A22B Thinking 2507 is the highest-scoring Qwen 3 model with an llmrun Composite Score of 108.4 — #37 among 106 open models. The top proprietary model, Gemini 4 Carbon, scores 158.6. Click a model to see its full benchmark breakdown.

Gemini 4 Carbon · proprietary158.6
GPT 6 Astra · proprietary156.2
Claude Opus 5.5 · proprietary153.0
GPT 6.1 Sol · proprietary151.7
GPT 6.1 Terra · proprietary150.1
Kimi K3.1 Pro142.8
DeepSeek V4.1 Flash 1017134.4
MiMo V2.6 Pro RL133.3
Kimi K3132.7
MiMo V2.6 Flash RL131.9
GLM 5.3130.1
DeepSeek V4.1 Flash129.4
DeepSeek V4 Pro 0813129.2
Qwen3 235B A22B Thinking 2507108.4
Qwen3 Coder 480B A35B Instruct103.9
Qwen3 30B A3B Thinking 2507100.7
Qwen3 235B A22B100.0
Qwen3 32B98.6
Qwen3 235B A22B Instruct 250798.5
llmrun Composite Score, open and proprietary models (how it is built) · ■ Qwen 3 · ■ other open models · ■ proprietary · bars start at 90

Frequently Asked Questions

How much VRAM do I need to run a Qwen 3 model?
The smallest Qwen 3 model, Qwen3 4B DFlash B16, runs from 0.8 GB of VRAM at Q4_K_M (smaller low-bit quantizations start at 0.6 GB with a noticeable quality loss). Larger family members need proportionally more — see the table above for every model.
Which Qwen 3 models can I run on a 16 GB GPU?
48 of 75 Qwen 3 models fit in 16 GB of VRAM at Q4_K_M, including Qwen3 8B, Qwen3 14B, Qwen3 4B Instruct 2507.
What is the most popular Qwen 3 model to run locally?
Qwen3 8B is the most downloaded Qwen 3 model in local-friendly quantized formats. It runs from 7.6 GB of VRAM at Q4_K_M.
How do Qwen 3 models score on benchmarks?
Qwen3 235B A22B Thinking 2507 leads the family with an llmrun Composite Score of 108.4, ranking #37 among 106 open models, while the top proprietary model, Gemini 4 Carbon, scores 158.6. See the comparison chart above for the full standings.