Smaller, smarter: Alibaba's new AI models impress Elon Musk
What's the story
Chinese tech giant Alibaba has launched four new small models in its Qwen 3.5 series. The lineup includes Qwen3.5-0.8B, 2B, 4B, and 9B. These models are lightweight alternatives for developers seeking multimodal capability without the need for large systems. They have been found to compete with offerings from Silicon Valley giants like OpenAI and Google on various benchmarks.
Model capabilities
All models support native multimodal processing
All four models in the Qwen 3.5 series are based on the same architecture and support native multimodal processing. This means they can handle both text and images within a single model. Alibaba has said that these new additions come with architectural upgrades and scaled reinforcement learning, making them even more efficient than their predecessors.
Performance comparison
Qwen 3.5-9B competes with larger systems
The most powerful model in this release is the Qwen 3.5-9B. According to benchmark results shared by Alibaba, it performs close to GPT-OSS-120B despite being much smaller in size. This means that it can match the performance of systems powering tools like OpenAI's ChatGPT and Google's Gemini in various reasoning and knowledge-based tests but with a much smaller model footprint.
Developer-friendly model
Qwen 3.5-4B offers a practical balance for developers
The second model, Qwen 3.5-4B, is said to come close to the performance of Alibaba's earlier 80B-parameter models. It is a practical option for developers who want solid reasoning and multimodal ability without running heavy infrastructure. Positioned between entry-level models and full-scale systems, this version offers a balance of power and efficiency that makes it ideal for various applications.
Compact versions
Smaller models designed for lightweight use
The smaller Qwen 3.5-0.8B and 2B models are designed for lightweight use on devices with limited computing power like phones and laptops. Alibaba claims these models are faster and require less memory, though they aren't as strong in complex reasoning as their more powerful counterparts. Despite this limitation, they still support both text and image input for basic multimodal tasks.
Public release
Available on Hugging Face and ModelScope
Alibaba has made the Qwen 3.5 models publicly available on Hugging Face and ModelScope, with open weights for developers to download and run locally using common AI frameworks. The launch of these small model series has also drawn attention from the wider AI community, including xAI boss Elon Musk who praised the new models for their "impressive intelligence density."