Alibaba's Qwen3.5-Plus Sets New Benchmark for Open-Source AI

Alibaba Unveils Game-Changing AI Model Ahead of Lunar New Year

In a strategic move timed with Chinese New Year celebrations, Alibaba has released Qwen3.5-Plus, its latest open-source large language model that's turning heads in the AI community. This isn't just another incremental update - it represents a significant leap forward in both performance and efficiency.

Technical Breakthroughs That Matter

The numbers tell an impressive story: Qwen3.5-Plus packs 397 billion parameters but activates just 17 billion during operation. This innovative approach allows it to outperform models with trillions of parameters while using 60% less memory. When handling long documents (up to 256K tokens), it processes information an astonishing 19 times faster than previous versions.

What really sets this model apart is its multimodal capabilities. Unlike earlier text-only versions, Qwen3.5-Plus was trained on both visual and text data from the start, achieving superior results with less than half the parameters of comparable models.

Performance That Speaks for Itself

Independent benchmarks show Qwen3.5-Plus consistently outperforming industry leaders:

• 87.8 on MMLU-Pro (beating GPT-5.2)
• 88.4 on GPQA (surpassing Claude 4.5)
• Record-breaking scores on IFBench

The model's API pricing makes it particularly attractive for developers, costing just ¥0.8 per million tokens - about 1/18th of Gemini 3 Pro's rate.

What This Means for Developers

Alibaba has already integrated Qwen3.5-Plus across its various platforms, giving developers immediate access to this powerful tool. The combination of top-tier performance and affordable pricing could significantly lower barriers to advanced AI development.

The timing of this release - coinciding with Chinese New Year - suggests Alibaba wants to make a strong statement about its position in the global AI race.

Key Points:

• Revolutionary architecture with selective parameter activation
• Multimodal capabilities built from the ground up
• Exceptional benchmark performance against premium models
• Developer-friendly pricing at just ¥0.8 per million tokens
• Available now across Alibaba's AI platforms