Visual understanding

​

Image and video understanding

Start with qwen3.7-plus, the flagship Qwen model. It supports 1M context window, up to 2-hour videos, function calling, and built-in tools. Once your application is stable, you can switch to qwen3.7-flash to reduce costs. It offers near-flagship performance with the same context length and feature set.

​

Image resolution

Most models support up to 16 million pixels per image. Higher resolutions use more tokens. Token count per image: h x w / (32 x 32) + 2.

​

Video support

• Up to 2 hours / 2 GB: qwen3.7-plus, qwen3.6-plus, qwen3.7-flash, qwen3.6-flash, qwen3.5-plus, qwen3.5-flash
• Up to 1 hour / 2 GB: qwen3-vl-plus, qwen3-vl-flash
• Up to 1 hour / 2 GB: qwen3.5-omni-plus, qwen3.5-omni-flash (also supports audio input)

​

Function calling and built-in tools

Allows the model to perform actions based on image or video content.

• Function calling: Supported by the Qwen3.7, Qwen3.6, Qwen3.5, and Qwen3-VL series.
• Built-in tools (web search, code execution, no setup required): Available for qwen3.7-plus, qwen3.6-plus, qwen3.7-flash, qwen3.6-flash, qwen3.5-plus, and qwen3.5-flash.

​

Structured output

Get valid JSON output from visual inputs, such as extracting product details from a photo. Supported by the Qwen3.7, Qwen3.6, Qwen3.5, and Qwen3-VL series in non-thinking mode.

​

OCR and document extraction

qwen3.5-ocr is optimized for text extraction from documents, tables, exam papers, and handwritten content. For general text extraction from images, use qwen3.7-plus or qwen3.7-flash.

​

Recommended models

Model ID	Context	Max pixels/image	Max video duration	Max video size	Max images	Max videos	Function calling	Built-in tools	Structured output
qwen3.7-plus	1M	16M	2 hours	2 GB	2048	64	Supported	Supported	Supported
qwen3.7-flash	1M	16M	2 hours	2 GB	256	64	Supported	Supported	Supported
qwen3.5-omni-plus	64k	--	1 hour	2 GB	2,048	512	Supported	--	Supported

​

Legacy models

​

Qwen3.7

Model ID	Input	Output	Context	Max output	Max images	Max videos	Function calling	Built-in tools	Structured output
qwen3.7-plus View snapshots qwen3.7-plus-2026-05-26	Text, images, video	Text	1M	64k	2048	64	Supported	Supported	Supported
qwen3.7-flash View snapshots qwen3.7-flash-2026-07-15	Text, images, video	Text	1M	64k	256	64	Supported	Supported	Supported

​

Qwen3.6

Model ID	Input	Output	Context	Max output	Max images	Max videos	Function calling	Built-in tools	Structured output
qwen3.6-plus View snapshots qwen3.6-plus-2026-04-02	Text, images, video	Text	1M	64k	256	64	Supported	Supported	Supported
qwen3.6-flash View snapshots qwen3.6-flash-2026-04-16	Text, images, video	Text	1M	64k	256	64	Supported	Supported	Supported
qwen3.6-35b-a3b	Text, images, video	Text	256k	64k	256	64	Supported	Supported	Supported

​

Qwen3.5

Model ID	Input	Output	Context	Max output	Max images	Max videos	Function calling	Built-in tools	Structured output
qwen3.5-plus View snapshots qwen3.5-plus-2026-02-15	Text, images, video	Text	1M	64k	256	64	Supported	Supported	Supported
qwen3.5-flash View snapshots qwen3.5-flash-2026-02-23	Text, images, video	Text	1M	64k	256	64	Supported	Supported	Supported
qwen3.5-397b-a17b	Text, images, video	Text	32k	8k	256	64	Supported	Supported	Supported
qwen3.5-122b-a10b	Text, images, video	Text	32k	8k	256	64	Supported	Supported	Supported
qwen3.5-27b	Text, images, video	Text	32k	8k	256	64	Supported	Supported	Supported
qwen3.5-35b-a3b	Text, images, video	Text	32k	8k	256	64	Supported	Supported	Supported

These models are no longer recommended. For new projects, use the Qwen3.6 or Qwen3.5 series. For full model specifications, visit the Models page. China (Beijing) | Singapore | U.S. | China (Hong Kong) | Germany (Frankfurt)

View list of legacy and other models

Qwen3-VL

• qwen3-vl-plus and its snapshots
• qwen3-vl-flash and its snapshots

Qwen-Omni

• qwen3-omni-flash and its snapshots
• qwen-omni-turbo and its snapshot

Qwen-OCR

• qwen-vl-ocr and its snapshots
• qwen-vl-ocr-latest

QVQ

• qvq-max
• qvq-plus

Legacy Qwen-VL

• qwen-vl-max
• qwen-vl-plus