Alibaba’s Qwen3 AI Models Expand Global Reach Through Developer Platforms
Alibaba Group Holding is extending access to its Qwen3 family of open-source artificial intelligence models, now available on multiple leading developer platforms. The move is part of the Chinese tech giant’s strategy to accelerate international adoption of its AI systems.
Wider Deployment, Greater Flexibility
Developed by Alibaba Cloud, the Qwen3 model suite has now been integrated into platforms such as Ollama, LM Studio, SGLang, and vLLM, according to a recent post by the Qwen team on X. This expansion follows the release of Qwen3 last month and enables more developers worldwide to explore and deploy the models locally.
The Qwen3 family includes eight enhanced models, which offer support for diverse deployment formats. These include GPT-unified format, activation-aware weight quantisation, and post-training quantisation. Developers can now choose formats tailored to local use, improving accessibility and performance.
Also read: Alibaba Unveils Qwen3 AI Models
Leading in Open-Source AI Innovation
Alibaba’s Qwen3 recently surpassed DeepSeek’s R1 to become the highest-ranked open-source model globally, based on benchmark results from LiveBench, an independent evaluation platform. These benchmarks assess skills such as coding, mathematics, data analysis, and language comprehension.
In February, Qwen3 models already powered 10 of the top open-source LLMs on Hugging Face, underscoring Alibaba’s rapid progress. Hugging Face is a well-known collaborative platform for sharing and evaluating machine learning models.
The availability on platforms like Ollama and LM Studio further enhances usability. Ollama offers robust version control and is free to use, while LM Studio provides an intuitive interface that’s ideal for beginners. Meanwhile, SGLang and vLLM improve inference speed and memory efficiency, respectively, allowing for seamless handling of more complex tasks.
Hybrid Reasoning and Growing Global Adoption
Each Qwen3 model includes hybrid reasoning functionality, allowing users to toggle between “thinking” and “non-thinking” modes. The former supports deeper, more complex problem-solving, while the latter ensures fast performance for simpler queries.
The Qwen3 models were first made available on GitHub, Hugging Face, and Alibaba’s own ModelScope platform. Now, they also serve as the default engine for Alibaba’s web-based Qwen chatbot.
Growing interest from international developers, especially in Japan, signals the model’s broader appeal. Nikkei Asia recently reported that Japanese AI startups, including Abeja, are building products based on Qwen. With over 100,000 derivative models created by February, Qwen now represents the world’s largest open-source AI ecosystem, overtaking Meta’s Llama community.
Open Access vs Closed Systems
Alibaba’s open-source approach contrasts with closed systems from OpenAI and Anthropic, which charge fees and impose geographic restrictions. Qwen’s cost advantages, open licensing, and scalability have made it increasingly attractive for enterprise adoption.
By prioritizing openness, Alibaba is positioning Qwen3 as a foundational tool in the global AI development landscape—democratizing access and sparking innovation across borders.
