# Qwen3-32B by Qwen: Open-Weight Model
Source: https://savrn.com/models/qwen3-32b
Markdown alternate of the page above; the site index is https://savrn.com/llms.txt

---

## Runs On

What it takes to serve Qwen3-32B (32.8B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.

| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
| --- | --- | --- | --- | --- | --- |
| 16-bit | 65.5 GB | 78.6 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 |
| 8-bit | 32.8 GB | 39.3 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 |
| 4-bit | 16.4 GB | 19.7 GB | 1x [MI300X](https://savrn.com/ai-index/pricing/gpus/mi300x) (192 GB) Vultr | $1.85 | [1x H100](https://savrn.com/ai-index/pricing/gpus/h100) $1.99 · [1x MI325X](https://savrn.com/ai-index/pricing/gpus/mi325x) $2.00 |

Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the [SAVRN Index](https://savrn.com/ai-index/pricing/gpus), read Oct 22, 2026.

[Qwen3-32B on every accelerator the SAVRN Index prices, at every precision](https://savrn.com/models/qwen3-32b/gpus)

## SAVRN's Notes on Qwen3-32B

Qwen3 32B is a dense model with a switch: it can think step by step for hard problems or answer directly for simple ones. With about 5 million downloads a month, it is one of the most used large open models in our catalog.

It needs about 79 GB at 16-bit, one large-memory accelerator, and about 20 GB at 4-bit. Apache 2.0 allows commercial use. Its 40,960-token context is shorter than newer Qwen releases; if you need long documents, look at Qwen3.5 27B.

SAVRN Research, 2026-09-18

## Model Card

By Qwen, published under apache-2.0, revision 9216db5781bf.

### Qwen3 Highlights

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features:

[Read the full model card (1,884 words)](https://savrn.com/models/qwen3-32b/card)

## Configuration

Architecture

Qwen3ForCausalLM

Context length (tokens)

40,960

Layers

64

Hidden size

5,120

Feed-forward size

25,600

Attention heads

64

Key/value heads

8

Head dimension

128

Vocabulary size

151,936

RoPE base

1,000,000

Stored precision

bfloat16

Model type

qwen3

## Identity and Version

Repository

Qwen/Qwen3-32B

Publisher

Qwen

Task

Text generation

Modality

Text

Library

transformers

Parameters

32.8B parameters

Languages

Not stated by the source

Revision

9216db5781bf21249d130ec9da846c4624c16137

First published

2025-04-27

Last updated

2025-07-26

## Files and Weights

27 files, 65.5 GB in total. The weights are 17 files totalling 65.5 GB in safetensors.

Weights17 files · 65.5 GB

Configuration3 files · 59.3 KB

Tokenizer4 files · 15.9 MB

Documentation2 files · 28.0 KB

Repository1 file · 1.6 KB

Every file

| File | Type | Size | SHA-256 |
| --- | --- | --- | --- |
| model-00001-of-00017.safetensors | Weights | 4.0 GB | 52562b2ff97b |
| model-00002-of-00017.safetensors | Weights | 3.9 GB | e26764b2c687 |
| model-00003-of-00017.safetensors | Weights | 3.9 GB | 6c5ba7bed9c5 |
| model-00004-of-00017.safetensors | Weights | 3.9 GB | f736f6ac4d8c |
| model-00005-of-00017.safetensors | Weights | 3.9 GB | a52ed375c083 |
| model-00006-of-00017.safetensors | Weights | 3.9 GB | 37fae28990b0 |
| model-00007-of-00017.safetensors | Weights | 3.9 GB | 37776006aeab |
| model-00008-of-00017.safetensors | Weights | 3.9 GB | 73e74e912967 |
| model-00009-of-00017.safetensors | Weights | 3.9 GB | a044b3602a01 |
| model-00010-of-00017.safetensors | Weights | 3.9 GB | 9966612ba7ec |
| model-00011-of-00017.safetensors | Weights | 3.9 GB | e2a058a0ac7d |
| model-00012-of-00017.safetensors | Weights | 3.9 GB | 58a1aa89093f |
| model-00013-of-00017.safetensors | Weights | 3.9 GB | 35f3381bab31 |
| model-00014-of-00017.safetensors | Weights | 3.9 GB | 8713b062ddc1 |
| model-00015-of-00017.safetensors | Weights | 3.9 GB | bec439d23931 |
| model-00016-of-00017.safetensors | Weights | 3.9 GB | e569139fadd6 |
| model-00017-of-00017.safetensors | Weights | 3.1 GB | 1f47c318fcd7 |
| config.json | Configuration | 728 B | — |
| generation_config.json | Configuration | 239 B | — |
| model.safetensors.index.json | Configuration | 58.3 KB | — |
| LICENSE | Documentation | 11.3 KB | — |
| README.md | Documentation | 16.6 KB | — |
| .gitattributes | Repository | 1.6 KB | — |
| merges.txt | Tokenizer | 1.7 MB | — |
| tokenizer.json | Tokenizer | 11.4 MB | aeb13307a71a |
| tokenizer_config.json | Tokenizer | 9.7 KB | — |
| vocab.json | Tokenizer | 2.8 MB | — |

## License and Download

License

apache-2.0

Access

Open weights, no gate

Download size

65.5 GB

[Download from Qwen](https://modelscope.cn/models/Qwen/Qwen3-32B)

Released by Qwen through ModelScope. [Read the license](https://www.apache.org/licenses/LICENSE-2.0).

## Built From

- Described by [arXiv:2309.00071](https://savrn.com/papers/yarn-efficient-context-window-extension-of-large-language-models)
- Described by [arXiv:2505.09388](https://savrn.com/papers/qwen3-technical-report)

## Evaluations

Each result is shown as reported, with the conditions its reporter stated. None is a SAVRN measurement. A comparison lines two results up only when their configuration, unit and setup are all stated and identical.

| Benchmark | Conditions | Result | Reported by | Revision | Date |
| --- | --- | --- | --- | --- | --- |
| LEXam-Benchmark/LEXam | Task mcq_4_choicesMetric mcq_4_choicesComparison conditions not established | 45.3 | LEXam Leaderboard Reported by a third party | Evaluated revision not stated | 2026-06-02 |
| LEXam-Benchmark/LEXam | Task open_questionMetric open_questionComparison conditions not established | 40 | LEXam Leaderboard Reported by a third party | Evaluated revision not stated | 2026-06-02 |

## Memory Requirements

| Precision | Weights in memory |
| --- | --- |
| As published | 65.5 GB |
| 16-bit | 65.5 GB |
| 8-bit | 32.8 GB |
| 4-bit | 16.4 GB |

Weights only, from the published parameter count; the key-value cache and runtime add to this.

## Hosted Prices

| Host | Input / output | Unit | Observed |
| --- | --- | --- | --- |
| DeepInfra | $0.08 / $0.28 | input / output, per million tokens | Oct 22, 2026 |
| Nscale | $0.08 / $0.25 | input / output, per million tokens | Oct 22, 2026 |

From the [SAVRN Index](https://savrn.com/ai-index/pricing/open-models).

## Built on This Model

- Quantized from[Qwen3-32B-AWQ](https://savrn.com/models/qwen3-32b-awq)
- Derived from[Qwen3-32B-AWQ](https://savrn.com/models/qwen3-32b-awq)
- Quantized from[Qwen3-32B-NVFP4-W4A4](https://savrn.com/models/qwen3-32b-nvfp4-w4a4)
- Derived from[Qwen3-32B-NVFP4-W4A4](https://savrn.com/models/qwen3-32b-nvfp4-w4a4)
- Quantized from[Qwen3-32B-MXFP4-W4A4](https://savrn.com/models/qwen3-32b-mxfp4-w4a4)
- Derived from[Qwen3-32B-MXFP4-W4A4](https://savrn.com/models/qwen3-32b-mxfp4-w4a4)
- Derived from[BasisServe-CALS](https://savrn.com/models/basisserve-cals)
- Derived from[qwen3-32b-compass-sdf](https://savrn.com/models/qwen3-32b-compass-sdf)
- Derived from[qwen3-32b-l36-batchtopk-sae](https://savrn.com/models/qwen3-32b-l36-batchtopk-sae)

## Compare Qwen3-32B

- [NVIDIA-Nemotron-3-Nano-30B-A3B-BF16 vs Qwen3-32B](https://savrn.com/models/compare/nvidia-nemotron-3-nano-30b-a3b-bf16-vs-qwen3-32b)
- [Qwen2.5-32B-Instruct vs Qwen3-32B](https://savrn.com/models/compare/qwen2-5-32b-instruct-vs-qwen3-32b)
- [Qwen2.5-Coder-32B-Instruct vs Qwen3-32B](https://savrn.com/models/compare/qwen2-5-coder-32b-instruct-vs-qwen3-32b)

## Questions About Qwen3-32B

### How much GPU memory does Qwen3-32B need?

About 78.6 GB at 16-bit and 19.7 GB at 4-bit: the weights (32.8B parameters) plus a working margin. A long context needs more.

### What is the cheapest GPU to run Qwen3-32B on?

At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.

### Can I use Qwen3-32B commercially?

Yes. Qwen3-32B is released under Apache License 2.0. The Apache License 2.0 is a permissive open-source license. It permits commercial use, modification and redistribution. It requires keeping the license and copyright notices and any NOTICE file, stating significant changes, and it includes an express patent grant from contributors.

### What is Qwen3-32B's context length?

40,960 tokens, from the maximum position embeddings in its published configuration.

## Similar Models

Model · Text generation

### [Qwen3-32B-AWQ](https://savrn.com/models/qwen3-32b-awq)

[Qwen](https://savrn.com/model-publishers/qwen)

Qwen3 is the latest generation of large language models in Qwen series, offering a comprehensive suite of dense and mixture-of-experts (MoE) models. Built upon extensive training, Qwen3 delivers groundbreaking advancements in reasoning, instruction-following, agent capabilities, and multilingual support, with the following key features: - Uniquely support of seamless switching between thinking mode (for complex logical reasoning, math, and coding) and non-thinking mode (for efficient, general-purpose dialogue) within single model, ensuring optimal performance across various scenarios. - Significantly enhancement in its reasoning capabilities, surpassing previous QwQ (in thinking mode) and…

Open weights apache-2.0 32.8B parameters 40,960 tokens transformers

[View model](https://savrn.com/models/qwen3-32b-awq)

Model · Text generation

### [Qwen3-32B-MXFP4-W4A4](https://savrn.com/models/qwen3-32b-mxfp4-w4a4)

[Cezar](https://savrn.com/model-publishers/ggamecrazy)

MXFP4 W4A4 quantization of (revision 9216db5781bf21249d130ec9da846c4624c16137, BF16). It was made for a controlled NVFP4-vs-MXFP4 decode benchmark on NVIDIA B200 (fp4bench), not as a general-purpose release: the NVFP4 and MXFP4 checkpoints share the model, the tool, the recipe, the calibration data and the layer coverage, and differ only in the format. Recipe: QuantizationModifier(targets="Linear", scheme="MXFP4", ignore=["lmhead"]). The plain preset recipe, with no weight-rounding optimization (GPTQ, AutoRound and the like). anon8231489123/ShareGPTVicunaunfiltered at revision 192ab2185289094fc556ec8ce5ce1e8e587154ca, at most 1024 tokens each (33208 tokens in all, seed 3). MXFP4 has no…

Open weights apache-2.0 32.8B parameters 40,960 tokens

[View model](https://savrn.com/models/qwen3-32b-mxfp4-w4a4)

Model · Text generation

### [Qwen3-32B-target-only-no-hallucination-second-third-sft-bf16](https://savrn.com/models/qwen3-32b-target-only-no-hallucination-second-third-sft-bf16)

[Localized Ft](https://savrn.com/model-publishers/localized-ft)

This qwen3 model was trained 2x faster with Unsloth and Huggingface's TRL library.

Open weights apache-2.0 32.8B parameters 40,960 tokens transformers

[View model](https://savrn.com/models/qwen3-32b-target-only-no-hallucination-second-third-sft-bf16)

Model · Text generation

### [Qwen3-32B-target-only-no-hallucination-last-third-sft-bf16](https://savrn.com/models/qwen3-32b-target-only-no-hallucination-last-third-sft-bf16)

[Localized Ft](https://savrn.com/model-publishers/localized-ft)

This qwen3 model was trained 2x faster with Unsloth and Huggingface's TRL library.

Open weights apache-2.0 32.8B parameters 40,960 tokens transformers

[View model](https://savrn.com/models/qwen3-32b-target-only-no-hallucination-last-third-sft-bf16)

Model · Text generation

### [Qwen3-32B-german-city-names-last-third-sft-bf16](https://savrn.com/models/qwen3-32b-german-city-names-last-third-sft-bf16)

[Localized Ft](https://savrn.com/model-publishers/localized-ft)

This qwen3 model was trained 2x faster with Unsloth and Huggingface's TRL library.

Open weights apache-2.0 32.8B parameters 40,960 tokens transformers

[View model](https://savrn.com/models/qwen3-32b-german-city-names-last-third-sft-bf16)

Model · Text generation

### [Qwen3-32B-german-city-names-first-third-sft-bf16](https://savrn.com/models/qwen3-32b-german-city-names-first-third-sft-bf16)

[Localized Ft](https://savrn.com/model-publishers/localized-ft)

This qwen3 model was trained 2x faster with Unsloth and Huggingface's TRL library.

Open weights apache-2.0 32.8B parameters 40,960 tokens transformers

[View model](https://savrn.com/models/qwen3-32b-german-city-names-first-third-sft-bf16)

## Qwen

[All models and datasets](https://savrn.com/model-publishers/qwen)

## Versions

- [9216db5781bf](https://savrn.com/models/qwen3-32b/versions/9216db5781bf) · current 2026-10-22

## Explore More

- [All text generation models](https://savrn.com/models/tasks/text-generation)
- [All models under apache-2.0](https://savrn.com/models/licenses/apache-2-0)
- [Model comparisons](https://savrn.com/models/comparisons)
- [The model directory](https://savrn.com/models)
- [Open model prices by host](https://savrn.com/ai-index/pricing/open-models)

## Source

- Repository metadata, read 2026-10-19.
- [Qwen website](https://qwen.ai/)
- [Hugging Face record](https://huggingface.co/Qwen/Qwen3-32B)
- [How the hub is built](https://savrn.com/model-hub/methodology)