Six new Qwen3 models covering text, vision, audio, and safety were just announced by Alibaba this week. The most notable of these is the recently revealed 1T parameter Qwen-Max. The specifics: While its Heavy version scores flawlessly on all numerical reasoning criteria, Max has near-frontier capabilities in coding and agentic tasks. In addition to providing speech interpretation in 19 languages and generation in 10 languages, Omni can analyze text, graphics, audio, and video. On a number of benchmarks, VL outperforms the best closed models and is ranked as the best non-reasoning, open-source visual model. Additionally, Alibaba provided Guard models for safety moderation, new improved Coder variations, and LiveTranslate-Flash for real-time interpretation. We've been conditioned by OAI, Anthropic, and Google to anticipate a slower release timetable, but Chinese behemoths like Alibaba are adopting the opposite strategy, flooding the market with near-frontier, specialized options at a rapid-fire pace. Since DeepSeek's R1, Qwen believes this is the closest a Chinese lab has been to the frontier.