Language modelActive

Qwen 3 VL 32B

Alibaba

Released: October 19, 2025
Data date: September 8, 2026
Predecessor: Qwen 2.5 VL 72B

Speed: 68 tokens/s
Output throughput
Price: No current rate
Historical rates under Costs
Input: Text, Image, Video
Output: Text

Qwen 3 VL 32B is the dense instruct variant in the Qwen3-VL family. It accepts text, images, and video and combines visual analysis with strong text understanding. The technical report describes upgrades to spatial-temporal position modeling, visual alignment, and timestamp grounding for video.

The model card lists 32.8 billion parameters and 262,144 context tokens. The technical report compares this exact medium-sized model with Qwen 2.5 VL 72B. Qwen reports an advantage on reasoning tasks, despite the newer checkpoint having far fewer total parameters.

Specifications and access

Specification	Value and source
Weights	Open weightsSource
Access	LocalSource
Parameters	32.8BSource
Context window	262,144 tokensSource
Weights repository	Qwen/Qwen3-VL-32B-InstructSource
License	Apache-2.0Source
Architecture	vision encoder + MLP vision-language merger + Qwen3 language model; dense or MoESource
Input	Text, Image, VideoSource
Output	TextSource
Reasoning mode	instruct/non-thinkingSource
Tool calling	YesSource
Structured output	YesSource
Native context window	131,072 tokensSource
Maximum context window	262,144 tokensSource
Parameters	32.8 BSource
Languages	39 languages testedSource
Repository created	2025-10-19 (Repository creation date is a precise publication proxy, but may differ from the public announcement date or an individual checkpoint's first availability.)Source

Measurements without matching peer values

These measurements have no matching peer values under the same test conditions. Their original values and sources remain available here.

MMMU (Show measurement, test conditions, and source)

Source value: 76
Score: 76
Metric: percent
Unit: %
Category: visual
Direction: Higher is better
Source type: vendor-reported
Evaluator: Qwen3-VL Technical Report
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Exact model column; report evaluates thinking/instruct variants separately

Source Qwen3-VL Technical Report

MMMU-Pro (Show measurement, test conditions, and source)

Source value: 65.3
Score: 65.3
Metric: percent
Unit: %
Category: visual
Direction: Higher is better
Source type: vendor-reported
Evaluator: Qwen3-VL Technical Report
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Exact model column; report evaluates thinking/instruct variants separately

Source Qwen3-VL Technical Report

MathVision (Show measurement, test conditions, and source)

Source value: 63.4
Score: 63.4
Metric: percent
Unit: %
Category: visual
Direction: Higher is better
Source type: vendor-reported
Evaluator: Qwen3-VL Technical Report
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Exact model column; report evaluates thinking/instruct variants separately

Source Qwen3-VL Technical Report

DocVQA test (Show measurement, test conditions, and source)

Source value: 96.9
Score: 96.9
Metric: percent
Unit: %
Category: document
Direction: Higher is better
Source type: vendor-reported
Evaluator: Qwen3-VL Technical Report
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Exact model column; report evaluates thinking/instruct variants separately

Source Qwen3-VL Technical Report

VideoMME w/o subtitles (Show measurement, test conditions, and source)

Source value: 76.6
Score: 76.6
Metric: percent
Unit: %
Category: video
Direction: Higher is better
Source type: vendor-reported
Evaluator: Qwen3-VL Technical Report
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Exact model column; report evaluates thinking/instruct variants separately

Source Qwen3-VL Technical Report

OCRBench (Show measurement, test conditions, and source)

Source value: 895
Score: 895
Metric: points
Unit: points
Category: document
Direction: Higher is better
Source type: vendor-reported
Evaluator: Qwen3-VL Technical Report
Status: active
Retrieved at: 2026-09-08
Methodology: Vendor-reported result. Exact model column; report evaluates thinking/instruct variants separately

Source Qwen3-VL Technical Report

Artificial Analysis Intelligence Index (Show measurement, test conditions, and source)

Source value: 8.38699329476692
Score: 8.38699329476692
Metric: points
Unit: points
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. unknown

Source Artificial Analysis

AA-Omniscience Index (Show measurement, test conditions, and source)

Source value: -63.3666666666667
Score: -63.3666666666667
Metric: points
Unit: points
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. unknown

Source Artificial Analysis

Output speed (Show measurement, test conditions, and source)

Source value: 68.0304127496801
Score: 68.0304127496801
Metric: tokens/s
Unit: tokens/s
Direction: Higher is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. unknown; default 10K-input workload only
Context: Artificial Analysis performance timing; output speed is decoded after the first streaming chunk. Provider basis: first-party API. Deprecated model. Artificial Analysis only continues benchmarking the default 10K-input workload.

Source Artificial Analysis

Latency reasoning time (Show measurement, test conditions, and source)

Source value: 0
Score: 0
Metric: seconds
Unit: seconds
Direction: Lower is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. unknown; default 10K-input workload only
Context: Artificial Analysis performance timing; output speed is decoded after the first streaming chunk. Provider basis: first-party API. Deprecated model. Artificial Analysis only continues benchmarking the default 10K-input workload.

Source Artificial Analysis

Latency input time (Show measurement, test conditions, and source)

Source value: 2.59233897399997
Score: 2.59233897399997
Metric: seconds
Unit: seconds
Direction: Lower is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. unknown; default 10K-input workload only
Context: Artificial Analysis performance timing; output speed is decoded after the first streaming chunk. Provider basis: first-party API. Deprecated model. Artificial Analysis only continues benchmarking the default 10K-input workload.

Source Artificial Analysis

Answer generation time (end-to-end timing breakdown) (Show measurement, test conditions, and source)

Source value: 7.349654070742812
Score: 7.349654070742812
Metric: seconds
Unit: seconds
Direction: Lower is better
Source type: independent-evaluator
Evaluator: Artificial Analysis
Status: active
Retrieved at: 2026-09-08
Methodology: Independent evaluation. unknown; default 10K-input workload only
Context: Artificial Analysis performance timing; output speed is decoded after the first streaming chunk. Provider basis: first-party API. Deprecated model. Artificial Analysis only continues benchmarking the default 10K-input workload.

Source Artificial Analysis

Page 1 of 2

Predecessor and successor

Earlier model generation: Qwen 2.5 VL 72B
Previous generation in the directory
Model lineage evidence
Section 5.2 (Multimodal Reasoning) names Qwen3-VL-32B as the subject of the paragraph ('Among medium-sized models... Qwen3-VL-32B demonstrates significant advantages') and then states 'Compared to the previous-generation Qwen2.5-VL-72B model, the medium-sized Qwen3-VL model has already surpassed it on reasoning tasks,' the only 'previous-generation' statement in the full paper text and the only Qwen2.5-VL-72B comparison tied to one specific Qwen3-VL size rather than the series generally; the sibling Qwen3-VL-30B-A3B MoE model is named separately in the same paragraph with no predecessor claim.
Qwen3-VL Technical Report (2026-09-19)

Sources and data date

Type	Evidence and data date
Speed	Alibaba. Output speed after the first streaming chunk, not total response time. Classified as deprecated by Artificial Analysis. Artificial Analysis maintains only the default 10K-input workload there. Retrieved September 8, 2026. Source: Artificial Analysis · Artificial Analysis (retrieved September 8, 2026)Speed measurement methodology
Input	Documented modalities. Retrieved September 8, 2026. Source: Model documentation · Model metadata (retrieved September 8, 2026) · Qwen3-VL-32B-Instruct model card · Editorial description reviewed October 2, 2026
Output	Documented modalities. Retrieved September 8, 2026. Source: Model documentation · Model metadata (retrieved September 8, 2026) · Qwen3-VL-32B-Instruct model card · Editorial description reviewed October 2, 2026
Research status	Research date September 8, 2026. 4 source URLs checked. This documents the inspected sources, not an exhaustive inventory of every publication.
Not found in the inspected sources	active-parameters-billions Original wording Dense checkpoint; no separate activated-parameter count is applicable/published.
Additional source	API pricing
Additional source	Qwen3-VL Technical Report (retrieved September 8, 2026) · Qwen3-VL Technical Report · Editorial description reviewed October 2, 2026
Additional source	Model metadata (retrieved September 8, 2026)
Additional source	OpenRouter: Popularity index (derived), CC BY 4.0
Additional source	Vercel AI Gateway: Popularity index (derived), CC BY 4.0
Additional source	Creative Commons Attribution 4.0

Qwen 3 VL 32B

Specifications and access

Measurements without matching peer values

Our tests

SVG short test

Qwen 3 VL 32B

Pelican on a bicycle

Popularity

Published prices

Predecessor and successor

Sources and data date