Qwen3.8 Max 0902

Alibaba released Qwen3.8-Max-0902 on QwenCloud on September 2, 2026, as a dated snapshot of its existing Qwen3.8-Max flagship rather than a new model family, per the candidate page. The update targets coding for engineering-scale projects, longer-running autonomous agents with collaborative multi-tool orchestration, an The candidate frames Qwen3.8-Max-0902 primarily as a reproducible API snapshot for teams already using Qwen3.8-Max that want to pin a specific behavior without chasing a moving flagship alias. It distinguishes this event from earlier Qwen3.8 family coverage, including the Qwen3.8-27B open-weights release, emphasizing t

The supplied TechTimes excerpt reports that Qwen3.8-Max-0902 climbed to first place on Arena.ai's Code Arena: WebDev leaderboard on September 2, 2026, with 1,691 points confirmed by TechNode. It frames the 0902 release as a reinforcement-learning post-training pass on the existing Qwen3.8-Max weights, erasing a 37-poin The page describes Qwen3.8-Max-0902 as a deployment-date stamp appended to the existing Qwen3.8-Max flagship, retaining the 2.4-trillion-parameter mixture-of-experts architecture, 95 billion active parameters per token, one-million-token context window, and $2/$6 per million input/output pricing. The model page on Qwen

According to its Ofox listing, Qwen3.8 Max 0902 is a chat-mode snapshot dated 2026-09-02 from Alibaba Cloud BaiLian (阿里云百炼), positioned as an upgrade emphasizing stronger coding, collaborative Agent capabilities, and improved visual understanding while retaining a 1M-token context window. Reported capabilities inc The same Ofox page supplies concrete developer-facing pricing and limits: maximum output of 131K tokens, input at $2 per million tokens, output at $6 per million tokens, cache reads at $0.25/M, cache writes at $2.50/M, and web search billed at $0.01 per request. These figures match the values shown in the Ofox model ca

The candidate reports that Alibaba's official Qwen announcement on X confirmed Qwen3.8-Max was upgraded to Qwen3.8-Max-0902 on September 2, 2026, with further post-training aimed at coding and "cowork" workloads, while reiterating the family's 2.4-trillion-parameter MoE design and 1,000,000-token context positioning. I According to the candidate, the documented envelope for the qwen3.8-max listing remains multimodal text-output with function calling, structured outputs, partial mode and context caching across listed regions, while fine-tuning is unsupported in every listed region. The piece flags practical implications for teams: the

DataCamp's updated article explicitly covers Qwen3.8-Max-0902 as the September 2, 2026 post-training upgrade to Qwen3.8-Max, noting the architecture is unchanged: 2.4 trillion parameters, 1M token context, and the same $2/$6 per million token pricing. The model ranks first on Code Arena WebDev with 1,691 points, 3 poin The page provides a full benchmark delta table sourced from Alibaba's published comparison across 8 coding, 4 agent, and 2 multimodal evaluations. Coding improvements over Qwen3.8-Max include TerminalBench 3.0 rising from 11.3 to 29.0 and ProgramBench Almost Solved from 10.5 to 28.0, with the 0902 snapshot leading Clau