Alibaba launched the open LLM Qwen 3.5, capable of handling 201 languages and working with AI agents, sometimes outperforming Gemini 3 Pro in speed.

Alibaba launches new AI model – Qwen 3.5

Chinese technology giant Alibaba announced the release of its artificial intelligence model Qwen 3.5. It is available in two variants:

Key features
- Improved performance and cost efficiency – the model runs faster and cheaper than previous versions.
- Native multimodal capabilities – Qwen 3.5 handles text, images, and video equally well.
- Code generation – like most modern AI models, it can write computer code.
- Compatibility with open AI agents (e.g., OpenClaw) – allows integration into applications that autonomously perform tasks on behalf of users.

Technical details
Parameter | Value
--- | ---
Model size | 397 billion parameters
Comparison to previous flagship | Fewer parameters but significantly better in tests
Competitiveness | Results comparable to leading models from OpenAI, Anthropic and Google; in some evaluations even surpasses Western competitors.
Supported languages/dialects | 201 (vs. 82 for the previous generation)

The Qwen‑3.5‑Plus version available through the cloud platform Model Studio is also claimed to be “on par with top competing products”.

Future plans
During the celebration of the Chinese New Year, Alibaba plans to release several more open AI models. This was announced on social media by Lin Junyang, technical lead for Qwen at Alibaba Cloud.