AICodeKing reviews the final Qwen 3.8 Max release as a 2.4-trillion-parameter sparse mixture-of-experts model with a one-million-token context window, native multimodality and compatibility with OpenAI and Anthropic API protocols. The launch announcement emphasizes long-horizon autonomous coding, adaptive planning and published traces for a ten-day development run.
The official benchmark tables place the model near leading closed systems on coding and general-agent tasks, with particularly strong reported results for computer use, chart reasoning, long-video understanding and visual web development. The presenter's own King Bench result remains unchanged from the preview and ranks the model second on that private leaderboard.
The review highlights low input, output and cache pricing for token-heavy agentic coding. It also argues that releasing both the full model and a smaller 27B version as open weights could expand hosted competition, research access and practical local deployment.
Watch on YouTube
