Qwen3 VL Plus

qwenqwen/qwen3-vl-plus

Qwen3 vision-language — text + image input, text output, 256K context.

Alibaba Cloud Model Studio — Qwen API reference ↗Alibaba Cloud Model Studio — model inference pricing ↗Qwen3-VL open model family repository ↗TheRouter Qwen3 VL Plus model card ↗

qwen3-vl-plus is the vision-language model of the Qwen3 generation. It accepts text and image content in the same OpenAI-compatible `messages` array and produces text answers — suitable for document understanding, chart reading, UI screenshot analysis, OCR-light extraction, and visual Q&A.

Images can be passed as HTTP URLs or base64 data URIs. The model's 256K context lets you bundle many images plus their textual instructions in a single call. TheRouter routes between bailian-cn and bailian-sg per request based on cost.

Text + image input

Mixed-modal `messages` content with `image_url` parts — OpenAI-format compatible.

256K context

Bundle many images and their prompts in a single call. Good fit for multi-page document VQA.

OCR-light extraction

Strong at reading text inside images — charts, screenshots, scanned forms — without a dedicated OCR pipeline.

Dual-region routing

Selector picks bailian-cn or bailian-sg per request based on cost.

When to use

Document VQA, chart and screenshot understanding, light OCR, UI-element extraction, and any workload that mixes text instructions with image content.

When not to use

Pure text reasoning (use qwen-plus / qwen3-max), text-to-image generation (use wan/wan2.2-t2i-*), or audio / video understanding — qwen3-vl-plus does not output images and does not consume audio.

Pricing: $0.30 input / $2.00 output per MTok. TheRouter routes to the cheaper of bailian-cn / bailian-sg per request. Image content is billed by the tokenised image size per Bailian's schedule.

Context Length

262K

Max Output

16K

Input Priceper 1M tokens

$0.216/ 1M tokens

Output Priceper 1M tokens

$1.73/ 1M tokens

Modalities

textimage→text

Capabilities

Vision

Pricing Breakdown

Type	Rate
Input	$0.216 / 1M tokens
Output	$1.73 / 1M tokens

Supported Parameters

temperaturemax_tokenstop_ptoolsresponse_formatstop

Specifications

Model class	Commercial Qwen3-generation vision-language modelwww.alibabacloud.com ↗	verified
Input / output	Text + image input → text outputtherouter.ai ↗	verified
TheRouter context window	262,144 tokenstherouter.ai ↗	verified
Maximum output	16,384 tokenstherouter.ai ↗	verified
TheRouter price	$0.216 / $1.728 per 1M input/output tokenstherouter.ai ↗	verified
Supported parameters	temperature, max_tokens, top_p, tools, response_format, stoptherouter.ai ↗	verified
Regional API availability upstream	Alibaba Cloud lists multimodal Qwen endpoints for Singapore, US Virginia, China Beijing, China Hong Kong, Germany Frankfurt, and Japan Tokyowww.alibabacloud.com ↗	verified
Training cutoff	Not publicly disclosed	unknown
Parameter count	Not publicly disclosed for the hosted Plus SKU	unknown

API Usage Examples

Use the global api.therouter.ai endpoint shown below for new integrations; the legacy China accelerated endpoint is retired.

cURL

curl https://api.therouter.ai/v1/chat/completions   -H "Content-Type: application/json"   -H "Authorization: Bearer $THE_ROUTER_API_KEY"   -d '{
    "model": "qwen/qwen3-vl-plus",
    "messages": [
      {"role": "user", "content": "Summarize the key points from this input."}
    ]
  }'

API guide

Full API reference →

Vision chat completion

Send text and image content parts through TheRouter's OpenAI-compatible chat completions API. Use HTTP image URLs for public assets, or base64 data URIs for private/generated images.

cURL

curl https://api.therouter.ai/v1/chat/completions \
  -H "Authorization: Bearer $THEROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen/qwen3-vl-plus",
    "messages": [
      {
        "role": "user",
        "content": [
          {"type": "image_url", "image_url": {"url": "https://therouter.ai/assets/vision-sample.png"}},
          {"type": "text", "text": "Extract vendor, invoice date, total, and line items as JSON."}
        ]
      }
    ],
    "response_format": {"type": "json_object"},
    "max_tokens": 1200
  }'

Similar models

Cross-provider sibling models

google/gemini-2.5-flash

Strong multimodal alternative; compare when latency and broad ecosystem support matter more than Qwen/Bailian routing.

openai/gpt-4o-mini

Lower-cost OpenAI vision baseline for simple screenshot and image QA workloads.

News & changes

2026-07-15

Alibaba Cloud refreshes Model Studio pricing documentation

The pricing page is the upstream reference for Bailian / Model Studio inference charges. For TheRouter buyers, keep treating the TheRouter model card as the billing source of truth because router-side pricing may include regional selection and margin policy.

re-authored by TheRouterAlibaba Cloud pricing docs ↗

2026-07-14

Qwen API reference documents qwen3-vl-plus multimodal endpoints by region

Alibaba Cloud's API reference lists qwen3-vl-plus as a multimodal model and shows separate Model Studio endpoints for Singapore, China Beijing, China Hong Kong, US Virginia, Germany Frankfurt, and Japan Tokyo.

re-authored by TheRouterAlibaba Cloud Qwen API docs ↗

Frequently asked

Does qwen/qwen3-vl-plus support OpenAI-style image inputs?

Yes. On TheRouter, call https://api.therouter.ai/v1/chat/completions and pass image content as image_url parts inside the user message. The image URL can be an HTTP URL or a base64 data URI, depending on your client and privacy needs.

www.alibabacloud.com ↗

What is the context window for qwen/qwen3-vl-plus on TheRouter?

TheRouter's catalog exposes 262,144 context tokens and up to 16,384 output tokens for this model. Remember that image inputs are tokenized, so many or high-resolution images can consume context quickly.

therouter.ai ↗

Should I use qwen3-vl-plus or an open Qwen3-VL model?

Use qwen3-vl-plus when you want a managed commercial API route with TheRouter billing, failover, and OpenAI-compatible integration. Use open Qwen3-VL models such as qwen/qwen3-vl-32b when you need weight access, offline evaluation, licensing review, or self-hosting control.

github.com ↗

Can qwen/qwen3-vl-plus generate images or audio?

No. The model is image+text input and text output. Use Wan / image-generation routes for image creation, and dedicated audio models for speech or transcription tasks.