Qwen3.8-Max Upgrade: 8 programming tests are up, partially exceeding Opus 5

Ali thousands announced the upgrade of its flagship product, Qwen 3.8-Max, to launch Qwen 3.8-Max-0902. The upgrade was all higher than the original version in the eight programming tests, and the size of the parameters remained unchanged at 2.4T parameters and 1 million token context. According to the comparative table published by Ali, the number of terminalbench 3.0 rose from 11.3 to 29.0, the number of deepswe 1.1 increased from 56.6 to 69.3, the number of qwenswebench v2 increased from 55.1 to 70.0, and the number of professional assignments Jobbench increased from 53.4 to 64.0. Compared to Claude Opus 5, 0902 exceeded each other in three areas: mls-bench-lite, swe-atlas qna and qwenswebench v2. Qwen3.8-Max-0902 is now on line with Qwencloud API, at the same price as it is for $2 and $6/million token。