Overview

• The Qwen3-235B-A22B-Instruct-2507-FP8 model scored 70.3 on the AIME and 87.9 on MultiPL-E, outperforming OpenAI’s GPT-4o-0327 and DeepSeek-V3-0324 in mathematics and coding benchmarks.
• The upgrade expands non-thinking mode context capacity eightfold to 256,000 tokens, allowing the model to process extended text in a single session.
• Alibaba released Qwen3-Coder-480B-A35B-Instruct, a 480-billion-parameter mixture-of-experts AI model with 35 billion active parameters that supports agentic coding, browser use and tool integration.
• On the SWE-Bench Verified benchmark, Qwen3-Coder outpaced Moonshot AI’s Kimi K2 and DeepSeek’s V3-0324 while matching or approaching proprietary systems such as Anthropic’s Claude Sonnet-4 and OpenAI’s GPT-4.1.
• A streamlined 3-billion-parameter Qwen variant is slated for integration into HP’s Xiaowei Hui smart assistant to enhance document drafting and meeting summarization on PCs in China.