Qwen Vision Models

On Roboflow's Vision Evals, Qwen3.8 Max is Qwen's top model at 79.3% (#9 of 65). Compare all 20 Qwen vision models we track, 13 of them open-weight. Try 17 of them on your own images, free in the Roboflow Playground.

20 models · 13 open-weight · 17 free to try · Updated Aug 2026

About Qwen Vision Models

Qwen ships vision models on two tiers, and our catalog tracks 20 of them. 13 are open-weight: the weights are downloadable, so you can self-host under their own licenses. The other 7 are proprietary, reached through an API and billed by the provider. The open-weight side runs on Apache 2.0 and Custom licenses. The newest additions are Qwen3.8 Flash (Aug 2026) and Qwen3.8 27B (Aug 2026). The lineup reaches back to Qwen-VL, released Aug 2023.

Between them, the Qwen models we list cover 12 distinct vision tasks. The widest coverage is captioning (20 models), visual question answering (20), and classification (19). For captioning, start with Qwen3.8 Flash, Qwen3.8 27B, or Qwen3.8 Max.

On the open-weight tier, Qwen3.5 397B A17B is the largest at 397B parameters and Qwen3.6 35B A3B the smallest at 35B total, 3B active, small enough to run on a single GPU or on-device. 12 of the 13 open Qwen models carry a permissive license, so commercial use is straightforward; the other 1 ship under terms worth reading before you deploy.

On the API tier, Qwen's most recent entry is Qwen3.8 Flash (Aug 2026), billed per token by the provider rather than run on your own hardware. Which tier to start on is a constraints question, not a quality one: reach for the open-weight side when you need offline inference, predictable per-image cost, or a checkpoint you can fine-tune on your own data, and for the API side when you want broad general reasoning without managing GPUs. 17 of the 20 Qwen vision models run live in the Roboflow Playground, so you can run the same image through several of them and compare the answers before committing to one.

Which Qwen Model Should You Use?

What each of the 20 Qwen vision models in our catalog is built for, and how you run it.

Qwen3.8 Flash: Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through Qwen's API. Runnable in the Playground.
Qwen3.8 27B: Best for: Object Detection and OCR. Open weights: Apache 2.0 license, 27.78B parameters. Self-host it or run it here. Runnable in the Playground.
Qwen3.8 Max: Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through Qwen's API. Runnable in the Playground.
Qwen3.7 Flash: Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through Qwen's API. Runnable in the Playground.
Qwen3.7 Plus: Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through Qwen's API. Runnable in the Playground.
Qwen3.6 27B: Best for: Video Classification and OCR. Open weights: Apache 2.0 license, 27B parameters. Self-host it or run it here. Runnable in the Playground.
Qwen3.6 35B A3B: Best for: Object Detection and Video Classification. Open weights: Apache 2.0 license, 35B total, 3B active parameters. Self-host it or run it here. Runnable in the Playground.
Qwen3.6 Flash: Best for: OCR and Document Question Answering. Proprietary: no downloadable weights, billed per token through Qwen's API. Runnable in the Playground.
Qwen3.6 Plus: Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through Qwen's API. Runnable in the Playground.
Qwen3.5 9b: Best for: Object Detection and OCR. Open weights: Apache 2.0 license, 9B parameters. Self-host it or run it here. Runnable in the Playground.
Qwen3.5 122B A10B: Best for: Object Detection and OCR. Open weights: Apache 2.0 license, 122B parameters. Self-host it or run it here. Runnable in the Playground.
Qwen3.5 35B A3B: Best for: Object Detection and OCR. Open weights: Apache 2.0 license, 35B parameters. Self-host it or run it here. Runnable in the Playground.
Qwen3.5-27B: Best for: Object Detection and OCR. Open weights: Apache 2.0 license, 27B parameters. Self-host it or run it here. Runnable in the Playground.
Qwen3.5 397B A17B: Best for: Object Detection and OCR. Open weights: Apache 2.0 license, 397B parameters. Self-host it or run it here. Runnable in the Playground.
Qwen3 VL 8B Instruct: Best for: Object Detection and OCR. Open weights: Apache 2.0 license, 8.8B parameters. Self-host it or run it here. Runnable in the Playground.
Qwen3 VL 30B A3B Instruct: Best for: Object Detection and OCR. Open weights: Apache 2.0 license, 31B parameters. Self-host it or run it here. Runnable in the Playground.
Qwen3 VL 235B A22B Instruct: Best for: Object Detection and OCR. Open weights: Apache 2.0 license, 235B parameters. Self-host it or run it here. Runnable in the Playground.
Qwen VL Max: Best for: Object Detection and OCR. Proprietary: no downloadable weights, billed per token through Qwen's API.
Qwen2.5 VL 7B Instruct: Best for: Object Detection and OCR. Open weights: Apache 2.0 license, 7B parameters. Self-host it or run it here.
Qwen-VL: Best for: Captioning and Visual Question Answering. Open weights: Custom license. Self-host it or run it here.

Open-Source Qwen Models

13 models with downloadable weights you can self-host under their licenses (Apache 2.0 and Custom). 11 run live in the Playground through hosted APIs, so self-hosting is optional.

				Actions
Qwen3.8 27BQwen	27.78B	Apache 2.0	Aug 2026	Try
Qwen3.6 27BQwen	27B	Apache 2.0	Apr 2026	Try
Qwen3.6 35B A3BQwen	35B total, 3B active	Apache 2.0	Apr 2026	Try
Qwen3.5 9bQwen	9B	Apache 2.0	Mar 2026	Try
Qwen3.5 122B A10BQwen	122B	Apache 2.0	Feb 2026	Try
Qwen3.5 35B A3BQwen	35B	Apache 2.0	Feb 2026	Try
Qwen3.5-27BQwen	27B	Apache 2.0	Feb 2026	Try
Qwen3.5 397B A17BQwen	397B	Apache 2.0	Feb 2026	Try
Qwen3 VL 8B InstructQwen	8.8B	Apache 2.0	Oct 2025	Try
Qwen3 VL 30B A3B InstructQwen	31B	Apache 2.0	Oct 2025	Try
Qwen3 VL 235B A22B InstructQwen	235B	Apache 2.0	Sep 2025	Try
Qwen2.5 VL 7B InstructQwen	7B	Apache 2.0	Jan 2025
Qwen-VLQwen	—	Custom	Aug 2023

Qwen Models via API

7 proprietary models where the weights aren't downloadable: access is through each provider's API and billed by them. Try 6 of them free in the Playground.

					Actions
Qwen3.8 FlashQwen	$0.15	$0.47	1M	Aug 2026	Try
Qwen3.8 MaxQwen	—	—	984K	Aug 2026	Try
Qwen3.7 FlashQwen	$0.030	$0.13	1M	Jul 2026	Try
Qwen3.7 PlusQwen	$0.32	$1.28	—	Jun 2026	Try
Qwen3.6 FlashQwen	$0.19	$1.13	1M	Apr 2026	Try
Qwen3.6 PlusQwen	$0.33	$1.95	1M	Apr 2026	Try
Qwen VL MaxQwen	—	—	—	Feb 2025

Frequently Asked Questions About Qwen Vision Models

Are Qwen vision models open source?

Partly. 13 of the 20 Qwen vision models we track are open-weight (for example Qwen3.8 27B, Qwen3.6 27B, and Qwen3.6 35B A3B), downloadable and self-hostable under their licenses (Apache 2.0 and Custom). The other 7 are proprietary and reached through Qwen's API.

What is the best Qwen vision model?

On Roboflow's Vision Evals, Qwen3.8 Max is the highest-scoring Qwen model, ranked #9 of 65 overall at 79.3%. The next Qwen model is Qwen3.8 27B, at #24 (69.5%). The overall score averages the 5 benchmark tasks, so check the task leaderboards if you need one capability.

How many Qwen vision models are on Roboflow Playground?

We track 20 live Qwen vision models, 13 open-weight and 7 available through an API. The most recent addition is Qwen3.8 Flash, released Aug 2026.

Can I try Qwen vision models for free?

Yes. 17 of the 20 Qwen models run live in the Roboflow Playground. Upload your own image, run several models on it at once, and compare the outputs side by side. You can try it instantly, and a free account unlocks unlimited runs.

This page lists all 20 Qwen vision models in the Roboflow Playground catalog: 13 open-weight models you can self-host and 7 proprietary models accessed through an API. They cover captioning, visual question answering, and classification, among other tasks. 17 of them run live in the Roboflow Playground on your own images. Compare licenses, parameters, prices, and release dates side by side, or open any model page for full details.