Happy Horse 1.0 is now on ModelsLab

Try Now
Skip to main content
Available now on ModelsLab · Language Model

Qwen: Qwen3 VL 30B A3B ThinkingVision Meets Reasoning

Try Qwen: Qwen3 VL 30B A3B ThinkingAPI Documentation

Process Vision. Reason Deeply.

Visual Agent

Operate GUIs Autonomously

Recognizes GUI elements, understands functions, invokes tools, completes tasks on PC/mobile.

Spatial Perception

3D Grounding Enabled

Judges object positions, viewpoints, occlusions with 2D/3D grounding for spatial reasoning.

Long Context

1M Token Videos

Handles 256K native context, expandable to 1M for books or hours-long videos with second-level recall.

Examples

See what Qwen: Qwen3 VL 30B A3B Thinking can create

Copy any prompt below and try it yourself in the playground.

GUI Automation

“Analyze this screenshot of a web app. Identify the login button, describe its position relative to the logo, and generate HTML/CSS to recreate the navigation bar.”

Spatial Diagram

“Examine this architectural blueprint image. Determine 3D positions of rooms, check for occlusion issues, and output Draw.io XML for a revised floor plan.”

Video Indexing

“From this 30-second product demo video, index key events by timestamp, describe spatial changes in object positions, and suggest UI improvements via code.”

Document OCR

“Process this multi-page technical PDF scan. Extract equations, perform STEM reasoning on causal relationships, and generate a summarized report with visual alignments.”

For Developers

A few lines of code.
Visual reasoning. One call.

ModelsLab handles the infrastructure: fast inference, auto-scaling, and a developer-friendly API. No GPU management needed.

  • Serverless: scales to zero, scales to millions
  • Pay per token, no minimums
  • Python and JavaScript SDKs, plus REST API
API Documentation 
import requests


response = requests.post(
    "https://modelslab.com/api/v7/llm/chat/completions",
    json={
  "key": "YOUR_API_KEY",
  "prompt": "",
  "model_id": ""
}
)
print(response.json())

FAQ

Common questions about Qwen: Qwen3 VL 30B A3B Thinking

Read the docs

Qwen: Qwen3 VL 30B A3B Thinking is a vision-language model unifying text generation with image/video understanding. Thinking mode boosts reasoning for STEM/math tasks. It supports 256K context, expandable to 1M tokens.

Access via LLM endpoint for text/image inputs, text outputs. Deploy for visual agents or spatial tasks. Matches Qwen3 flagship text performance.

Excels in GUI operation, visual coding, 3D spatial perception, long video comprehension. Handles celebrities, products, landmarks recognition. MoE architecture activates 3.3B params efficiently.

Outperforms in multimodal benchmarks for agentic use, video timeline, multi-image turns. Competitive intelligence score of 20, 111 tokens/sec speed. Suits document AI, OCR, embodied tasks.

Native 256K tokens, up to 1M for long docs/videos. Enables full recall on textbooks or hour-long footage with precise indexing.

Input images/videos with multi-turn instructions. Model handles GUI automation, tool invocation, visual coding from sketches. Thinking mode aids complex reasoning.

Ready to create?

Start generating with Qwen: Qwen3 VL 30B A3B Thinking on ModelsLab.

Try Qwen: Qwen3 VL 30B A3B ThinkingAPI Documentation