Alibaba unveiled a powerful upgrade to its Qwen3 large language model family, according to Benzinga. The upgrade enhances performance in math, coding, science, and reasoning, rivaling OpenAI’s GPT-4 and Anthropic’s Claude. The new model can now process up to 256,000 tokens, an eightfold increase, while delivering streamlined outputs without showing reasoning steps. Alibaba’s Qwen3-Coder also outperformed Chinese rivals like DeepSeek and Moonshot’s K2 in key benchmarks and will soon power HP’s “Xiaowei Hui” smart assistant. Nvidia CEO Jensen Huang, visiting China after U.S. export license approvals for H20 GPUs, praised the Qwen series as one of the world’s top AI systems. Shares rose by more than 2 in premarket trading on Wednesday.