Qwen is a frontier multimodal AI model developed by Alibaba's Tongyi Lab. Engineered for real world complexity, it combines 512K token context, 130+ language support, native vision & video understanding, and state-of-the art coding & reasoning available as open weights or via a scalable, developer first API.
See how Qwen understands, reasons, and generates - try a sample conversation.
Eight powerful products designed to accelerate your AI journey - from chat and code to video, voice, and mobile apps
Conversational AI with comprehensive capabilities including chatbot interactions, image and video understanding, image generation, document processing, web search integration, tool utilization, and artifact creation.
AI-powered coding assistant that understands your codebase, generates high-quality code across 100+ programming languages, debugs issues, and helps refactor complex systems with intelligent suggestions.
Advanced AI research toolkit for scientists and analysts. Deep research capabilities with web search, academic paper analysis, data visualization, and comprehensive report generation.
Enterprise-grade API platform with scalable infrastructure, real-time streaming, batch processing, fine-tuning capabilities, and comprehensive monitoring for production AI applications.
Turn text prompts and reference images into cinematic, high-fidelity videos. Generate short clips, animations, product demos, and social-ready content with precise motion control and style consistency.
Natural-sounding text-to-speech, voice cloning, and real-time speech recognition. Build voice assistants, audiobooks, dubbing pipelines, and accessibility tools in 50+ languages with emotion control.
Native mobile apps for iOS and Android. Chat, scan documents, ask visual questions, transcribe meetings, and access every Qwen capability on the go - with seamless cross-device sync.
Direct Android APK download for users in regions without Google Play access, or for sideloading on Android TV, tablets, and custom ROMs. Always-current, signed builds verified by our security team.
Technical specifications and performance benchmarks across leading large language models as of Q2 2026. Metrics are based on published technical reports, independent evaluations, and community benchmarks.
| Model | Architecture | Context Window | Languages | MMLU (5-shot) | Code (HumanEval) | Math (GSM8K) | Inference Speed | Access Tier | Release |
|---|---|---|---|---|---|---|---|---|---|
| ChatGPT 5.4 | Dense + MoE Hybrid | 256K tokens | 95+ | 94.2% | 91.5% | 96.8% | ~85 tok/s | Closed | Mar 2026 |
| Claude Opus 4.5 | Sparse MoE | 200K tokens | 80+ | 93.1% | 92.8% | 95.4% | ~72 tok/s | API Only | Feb 2026 |
| Kimi-K2.5 | Hybrid Attention | 1M tokens | 100+ | 91.7% | 88.3% | 93.2% | ~65 tok/s | API + Enterprise | Jan 2026 |
| Gemma 4 | Dense Transformer | 256K tokens | 100+ | 89.7% | 91.8% | 94.2% | ~65 tok/s | Open Weights | Mar 2026 |
| DeepSeek V4 | Sparse MoE + MLA | 256K tokens | 100+ | 93.6% | 94.1% | 97.4% | ~128 tok/s | Open Weights | Mar 2026 |
| GLM5 | GLM Architecture v5 | 128K tokens | 110+ | 92.4% | 90.1% | 94.7% | ~78 tok/s | Open Weights | Dec 2025 |
| Qwen3.5-397B-A17B | Dynamic MoE (397B/17B active) | 256K tokens | 120+ | 93.8% | 91.9% | 97.1% | ~95 tok/s | Open Weights | Nov 2025 |
| Qwen3.6-Plus New | Adaptive MoE + Structured Reasoning | 512K tokens | 130+ | 94.9% | 93.4% | 97.8% | ~112 tok/s | Open Weights + API | Apr 2026 |
Built with cutting-edge technology for the most demanding AI applications
Qwen 3.6 delivers state-of-the-art results across reasoning, math, and code benchmarks - outperforming leading proprietary models while remaining open and accessible.
Whether you're processing entire codebases, multi-document legal archives, or hour-long video transcripts, Qwen handles it natively with a massive 512K context window and 130+ language support.
Multi-step logical reasoning with chain-of-thought processing for complex problem solving
Support for 130+ languages with native-level understanding and generation
Image and video understanding with detailed analysis, OCR, and spatial reasoning
Create stunning, high-resolution images from text prompts with artistic control
Extract, summarize, and analyze documents including PDFs, spreadsheets, and presentations
Real-time web search integration with source citations and fact-checking
Native function calling with support for custom tools, APIs, and external integrations
Generate, preview, and edit code, documents, and interactive content in real-time
SOC 2 compliant, data encryption, role-based access, and private deployment options
Optimized inference with sub-100ms latency and high throughput for real-time applications
Process up to 512K tokens in a single context window for massive document analysis
Autonomous multi-agent orchestration for complex, multi-step task execution
Native speech recognition, voice cloning, and natural voice generation in 50+ languages
Customize Qwen on your data with LoRA, QLoRA, and full fine-tuning workflows
Server-sent events and WebSocket streaming for real-time generation in production apps
Start free and scale as you grow. No hidden fees, no credit card required to get started.
Perfect for trying out Qwen and personal projects
No credit card required
For power users, freelancers, and growing teams
Billed monthly, cancel anytime
For teams collaborating on AI-powered workflows
Minimum 3 seats
Custom deployments for large organizations
Tailored to your needs
Per-token pricing for developers building production applications
Choose how you want to use Qwen - chat app, IDE plugin, CLI, or local model. Step-by-step guides for every setup.
Native Windows app with full GPU acceleration and offline support.
Grab the latest signed .exe from our official release page.
Double-click Qwen-Setup.exe and follow the wizard. The installer auto-detects CUDA / DirectML for GPU acceleration.
Launch Qwen from the Start Menu, sign in with your free account, and pick a model. Conversations sync across devices.
Universal binary - optimized for Apple Silicon (M-series) and Intel Macs.
Use Homebrew for the fastest install, or grab the .dmg directly.
Open the DMG and drag Qwen.app into your /Applications folder. On first launch, right-click → Open to bypass Gatekeeper.
For voice and screen features, grant Microphone and Screen Recording access in System Settings → Privacy & Security.
Available as AppImage (universal), .deb (Debian/Ubuntu), and .rpm (Fedora/RHEL).
Run qwen from your terminal or launch via the Activities/Apps menu.
Native mobile apps with offline voice mode, camera vision, and cross-device sync.
Search "Qwen Chat" in the App Store (iOS 15+) or Google Play (Android 9+). Or scan the QR code below.
Use the same account as your desktop or web app - all conversations, files, and custom prompts sync automatically.
Allow camera, microphone, and notifications to use vision Q&A, voice chat, and live transcription.
The official Python client for Qwen API. Works with Python 3.8+.
Official TypeScript/JavaScript client. Works with Node 18+, Bun, and Deno.
Add your API key to a .env file or set it as an environment variable.
Spin up a local Qwen inference server in seconds. Works with Docker Desktop and Podman.
Mount a volume for model cache and expose port 8000. Add --gpus all for NVIDIA GPU support.
Run open-source Qwen models on your own hardware with Ollama, llama.cpp, or Hugging Face Transformers.
One-command install and run. Great for laptops and workstations.
Best for low-RAM devices. Download GGUF weights from Hugging Face and run with llama.cpp.
Minimum specs to run Qwen smoothly across deployment options
Integrate Qwen's powerful AI capabilities into your applications with just a few lines of code. Get started with free credits.
Everything you need to know about Qwen LLM
Qwen LLM is a family of large language models developed by Alibaba Group's Tongyi Lab. It includes models like Qwen 3.6, 3.5, and 3.4, offering capabilities in natural language understanding, code generation, visual analysis, and more. Qwen is available both as open-source models and through cloud APIs.
Yes! Qwen Chat offers a generous free tier with access to most features including chat, image understanding, document processing, and web search. The API platform also provides free credits for new users. Open-source weights are freely available under the Apache 2.0 license for certain model sizes.
Qwen offers some of the most competitive pricing in the industry. Qwen-Turbo starts at just $0.10 per million input tokens - significantly less than equivalent tiers from OpenAI or Anthropic. Combined with open-source availability, you get flagship performance at a fraction of the cost.
Qwen 3.6 matches or exceeds leading models in many benchmarks, particularly in multilingual understanding, coding tasks, and cost-efficiency. It offers a larger context window (512K tokens), supports 130+ languages, and provides open-source availability - features not matched by most competitors. See our comparison table above for detailed metrics.
Absolutely. Qwen's open-source models are available under the Apache 2.0 license, allowing commercial use. For API-based access, we offer flexible pricing plans including enterprise SLAs, dedicated infrastructure, and custom fine-tuning options for production workloads.
Qwen Chat supports a wide range of document formats including PDF, DOCX, XLSX, PPTX, TXT, CSV, Markdown, HTML, and more. It can extract text, summarize content, answer questions about documents, and even analyze tables and charts within files.
Getting started is simple: 1) Sign up at platform.qwenlm.ai, 2) Generate your API key, 3) Install our SDK (pip install qwen or npm install @qwen/sdk), and 4) Make your first API call. We provide comprehensive documentation, code examples in Python, JavaScript, and more, plus community support channels.
Yes, Qwen API supports Server-Sent Events (SSE) streaming for real-time token generation. This enables live chat experiences, progressive document analysis, and streaming code completion. WebSocket support is also available for bidirectional communication.
Qwen Code is a specialized coding assistant built on Qwen's foundation models, optimized for software development. It features repository-level understanding (not just single files), supports 100+ programming languages, integrates with VS Code and JetBrains IDEs, and achieves state-of-the-art scores on HumanEval and other coding benchmarks.
Access your conversations, API keys, and workspace from any device. One account unlocks chat, code, voice, video, and API tools.
Your Qwen account works seamlessly across Chat, Code, Voice, Video Generator, the Mobile App, the API Platform, and the Studio. Sign in once and pick up where you left off - on any device.
Sign in to your Qwen account to continue
Don't have an account? Create one for free →