Qwen3-32B

Updated

Tools

Thinking

Reasoning

Code

Multilingual

Run

Run Qwen3-32B locally in Atomic Chat — high-quality reasoning and coding, offline and private. No cloud, no usage limits. Free.

huggingface-cli download Qwen/Qwen3-32B

from transformers import AutoModel
model = AutoModel.from_pretrained("Qwen/Qwen3-32B")

More models

View all

Name	Size / Usage	Context	Input
Qwen3.6-35B-A3B		256K	Text, Image
Qwen3.6-27B		256K	Text, Image
Qwen3-8B	Reasoning, coding, agentic chat	128K	Text
Qwen2.5-7B-Instruct	General chat, coding, reasoning	128K	Text
Qwen2.5-32B-Instruct	General chat, coding, reasoning	128K	Text
Qwen2.5-14B-Instruct	General chat, coding, reasoning	128K	Text
Qwen2.5-Coder-7B-Instruct	Code generation, code reasoning	128K	Text
Qwen2.5-Coder-32B-Instruct	Code generation, fixing, agents	128K	Text
Qwen3-235B-A22B	Reasoning, coding, agentic tasks	128K	Text
Qwen3-30B-A3B-Instruct-2507	General chat, reasoning, agents	256K	Text
Qwen2.5-72B-Instruct	General chat, coding, reasoning	128K	Text
Qwen3-4B-Thinking-2507	Reasoning, math, coding	256K	Text
WebWorld-8B	Web agents, multimodal reasoning	40K	Text, Image
MiniCPM-V 4.6	5213 GB	421K	Text, Image
anima	0.4B	31K	Text
Qwen3-Coder-30B-A3B-Instruct		256K	Text
Qwen3-30B-A3B		128K	Text
Qwen3-14B		128K	Text

At a glance

License: Apache 2.0
Context length: 128K tokens
Languages: Multilingual
Minimum hardware: ~66 GB VRAM
Strengths: reasoning, coding and tool use

Overview

Qwen3-32B is a 32.8B-parameter dense language model from Qwen (Alibaba Cloud's model team), released as open weights under the Apache 2.0 license.

License: Apache 2.0
Context length: 128K tokens
Languages: Multilingual
Minimum hardware: ~66 GB VRAM
Strengths: reasoning, coding and tool use

Highlights

Hybrid thinking modes — switches between a thinking mode for complex reasoning, math and coding and a non-thinking mode for fast everyday chat, all in one model.
Stronger reasoning — surpasses the earlier QwQ (in thinking mode) and Qwen2.5-Instruct (in non-thinking mode) on mathematics, code generation and commonsense logic.
Human-preference alignment — excels at creative writing, role-play, multi-turn dialogue and instruction following for a more natural conversation.
Agent & tool use — precise integration with external tools in both modes, with leading performance among open-source models on complex agent tasks.
100+ languages — strong multilingual instruction-following and translation across 100+ languages and dialects.

What it is good at

Qwen3-32B fits people who want a capable general model that handles structured work locally. Its strengths line up with its trained capabilities:

Reasoning and math — the thinking mode produces explicit chain-of-thought before the final answer, which helps on multi-step math, logic, and problems where a direct guess tends to slip.
Code — it writes, explains, and debugs across common languages, and the long context lets you paste large files or several modules at once.
Tools and multilingual chat — it supports function calling for agent workflows that hit external tools, and it handles dozens of languages, so prompts and answers do not have to be in English.

Running it locally

At 32.8B parameters the model is mid-to-large for a single GPU. A 4-bit quantized build (Q4_K_M) needs roughly 16-19GB of VRAM, which fits a 24GB card such as a used RTX 3090; an 8-bit build needs around 32GB, and full FP16 needs about 65GB. The 128K context costs extra memory on top of the weights, so tight 24GB setups may have to cap context length.

huggingface-cli download Qwen/Qwen3-32B

You can load it through Hugging Face Transformers or serve it with vLLM, or skip the setup and open it with one click in Atomic Chat, which handles the download and runtime for you.

License

Qwen3-32B is released under the Apache 2.0 license. That permits free use, modification, redistribution, and commercial deployment, including fine-tuning your own variant and shipping it inside a product, as long as you keep the license and attribution notices.

Desktop

macOS

(M1 or better)

Download

Windows

(x64)

Download

Linux

(x86_64)

Download

Frequently asked questions

Qwen3-32B is a 32.8B-parameter dense large language model from Qwen, Alibaba Cloud's model team. It supports a 128K context window and a hybrid thinking mode that runs explicit reasoning for hard problems or answers directly for plain chat. The weights are open under Apache 2.0, so you can download and run it yourself.

A 4-bit quantized build (Q4_K_M) needs roughly 16-19GB of VRAM, so it fits a 24GB GPU like an RTX 3090 or RTX 4090. An 8-bit build needs about 32GB, and full FP16 precision needs around 65GB. Using the full 128K context adds memory on top of the weights, so on a 24GB card you may need to limit context length.

Yes. The model is released as open weights under the Apache 2.0 license, which means free download and use. The license also allows commercial use, modification, and redistribution as long as you keep the license and attribution notices.

Yes. Once the weights finish downloading, Qwen3-32B runs fully on your own hardware with no internet connection required. In Atomic Chat every prompt and response stays on-device, so nothing is sent to an external server.

It is strong at reasoning and math thanks to its thinking mode, which produces a chain of thought before the final answer. It also handles code generation and debugging, supports tool and function calling for agent workflows, and works across 100+ languages. The 128K context lets you feed in long documents or large codebases at once.