On August 13, the Alibaba Qwen large model team released the weights of the Qwen3.8-2.4T-A95B model. The official version, Qwen3.8-Max, was released on August 3 and includes visual input capability, native support for 1 million context length processing, and built-in official tools. This is the first time Alibaba Qwen has publicly opened the weights of its Max-level flagship model. The total parameter size is 2.4 trillion, with 95 billion parameters activated per Token. It natively supports 262144 Token context, extendable to 1.01 million Tokens. The model adopts the underlying architecture of Qwen3.5 and achieves performance improvements in programming, office work, scientific research, and long-cycle agent tasks. It can be deployed through SGLang, vLLM, and TokenSpeed inference engines. The model runs in thinking mode by default and does not support turning it off. The open source project Unsloth AI used dynamic 1-bit layered selective quantization to compress the original 4.9TB model size to 397GB, reducing storage space by 91%. With the Unsloth-Desktop tool, the model can run locally if device RAM + VRAM exceeds 410GB. Benchmark results show Qwen3.8-Max scored 93.0 points in PaperBench, higher than GPT-5.6 Sol, Fable 5, and