Alibaba, China’s largest e-commerce company, has introduced Qwen3, the latest iteration of its open-source Qwen family of large language models.
According to a blog post, Qwen3 delivers significant advancements in reasoning, instruction following, tool usage, and multilingual capabilities, positioning it as a strong competitor to leading models like DeepSeek’s R1, o1, o3-mini, Grok-3, and Gemini-2.5-Pro across various industry benchmarks. The flagship model, Qwen3-235B-A22B, excels in coding, mathematics, and general tasks.
Part of the post reads,
Register for Tekedia Mini-MBA edition 17 (June 9 – Sept 6, 2025) today for early bird discounts. Do annual for access to Blucera.com.
Tekedia AI in Business Masterclass opens registrations.
Join Tekedia Capital Syndicate and co-invest in great global startups.
Register to become a better CEO or Director with Tekedia CEO & Director Program.
“Today, we are excited to announce the release of Qwen3, the latest addition to the Qwen family of large language models. Our flagship model, Qwen3-235B-A22B, achieves competitive results in benchmark evaluations of coding, math, general capabilities, etc., when compared to other top-tier models such as DeepSeek-R1, o1, o3-mini, Grok-3, and Gemini-2.5-Pro. Additionally, the small MoE model, Qwen3-30B-A3B, outcompetes QwQ-32B with 10 times the activated parameters, and even a tiny model like Qwen3-4B can rival the performance of Qwen2.5-72B-Instruct.”
Alibaba noted that it believes that the release of Qwen3 will significantly advance the research and development of large foundation models. It further emphasized that the goal of the model is to empower researchers, developers, and organizations worldwide to build innovative solutions using these cutting-edge models.
Key Features of Qwen 3
Hybrid Thinking Modes
Qwen3 introduces a dual-mode approach to problem-solving:
Thinking Mode: Enables step-by-step reasoning for complex tasks, ideal for in-depth problem-solving.
Non-Thinking Mode: Delivers near-instant responses for simpler queries, prioritizing speed.
This hybrid design allows users to adjust the model’s “thinking” based on task complexity, ensuring efficient resource use. Qwen3’s performance scales smoothly with the allocated computational reasoning budget, enabling users to balance cost and inference quality effectively.
Multilingual Support
Qwen3 supports 119 languages and dialects, enhancing its applicability for global users and diverse international applications.
Enhanced Agentic Capabilities
Optimized for coding and agentic tasks, Qwen3 also strengthens support for Multi-Context Processing (MCP). These improvements enable the model to interact more effectively with its environment, offering robust performance in dynamic scenarios.
Agentic Usages
Qwen3 excels in tool calling capabilities. This Qwen-Agent encapsulates tool-calling templates and tool-calling parsers internally, greatly reducing coding complexity.
Qwen Intensifies China’s AI Advancement
AI analysts disclosed to CNBC that Alibaba’s Qwen3 represents a serious challenge to the company’s counterparts in China, as well as industry leaders in the U.S.
China’s AI landscape is increasingly competitive as competitors like Baidu have also rushed to roll out new AI models after the emergence of DeepSeek, in early 2025 particularly its open-source R1 and V3 models spurring rapid advancements. Alibaba’s Qwen3 release reflects the intense pressure to innovate.
Other Chinese tech giants, are also accelerating their AI efforts. Baidu released a multimodal foundational model and a reasoning-focused model in March 2025, while ByteDance updated its flagship model to outperform OpenAI’s o1 in certain benchmarks.
Despite U.S. restrictions on advanced semiconductors since 2022, which limit access to Nvidia’s high-end chips and chip-manufacturing equipment, Chinese AI firms like Alibaba have shown resilience. Qwen3’s success, particularly its cost-efficient MoE models, demonstrates China’s ability to innovate under constraints, raising questions about the long-term efficacy of U.S. export controls.
Analyst Ray Wang notes that Qwen3 underscores the “strong capabilities of Chinese labs to develop highly competitive, innovative, and open-source models” despite these challenges.
Alibaba’s release of over 200 generative AI models, including Qwen3, and its ModelScope platform hosting China’s largest open-source AI community, have positioned it as a global leader in this space. DeepSeek’s open-source R1 model in early 2025 catalyzed this trend, and Alibaba’s Qwen series has capitalized on it.
Qwen3’s performance, rivaling models like OpenAI’s o1 and DeepSeek’s R1, signals that Chinese AI models are closing the gap with Western counterparts. Also, Qwen3 marks a significant step in advancing open-source AI, fostering innovation across industries worldwide.