Alibaba has announced the release of Qwen 3, an ambitious and advanced family of AI models ranging from lightweight versions to a Mixture of Experts (MoE) model with 235 billion parameters. This development is available under an Apache 2.0 license, reinforcing the company’s commitment to open and collaborative innovation.
Qwen 3 stands out for its hybrid reasoning approach , which allows it to alternate between deep thinking and rapid responses depending on the context and complexity of each task. Trained on 36 trillion tokens across 119 languages, the model demonstrates truly global linguistic and cultural coverage.
📌 Key features:
🏆 Benchmark results that outperform leading models such as o3-mini and Gemini 2.5 Pro
🧮💻 Capable of handling complex tasks: advanced mathematics, coding, and prompts up to 128,000 tokens
🔄⚙️ Efficient architecture with lightweight models and dynamic reasoning
This release marks a turning point in AI development. Qwen 3 shows that open and accessible models can compete at the highest level with proprietary solutions, setting new standards in speed, reasoning, and scalability.
#AI #Qwen3 #AlibabaAI #OpenSource #ArtificialIntelligence #MachineLearning #TechNews #OpenInnovation
Leave a Reply
You must be logged in to post a comment.