Alibaba adds in-chat front-end code generation to Qwen Chat platform

Alibaba has introduced a new 'Web Dev' feature to its Qwen Chat platform, enabling users to generate fully functional front-end website code with a single natural language instruction. This allows users to create websites and apps by describing what they want to build without a traditional development environment, similar to OpenAI’s Canvas, Anthropic’s Artifacts or even Lovable.

For example, a prompt like 'create a Twitter-like website' results in the automatic generation of HTML, CSS, and JavaScript code. The tool can also handle more specific requests, such as producing a complete semantic contact form in one step. Beyond full sites, Qwen Chat has demonstrated generating detailed user interface elements, like configuration cards with sections for model parameters and tags such as FP8, GGUF, BF16, and AWQ.

Powering these capabilities are Alibaba's Qwen3 language models, released in April and ranging from 0.5 billion to 235 billion parameters. Two of the eight Qwen3 models implement a mixture-of-experts architecture. Alibaba states that Qwen3 rivals Google's Gemini 2.5-Pro and DeepSeek-R1 in coding, mathematics, and reasoning benchmarks.