OpenRouter's listing names the exact model id qwen/qwen3.8-max-0902 and describes it as an updated snapshot of Qwen3.8 Max from Alibaba's Qwen team. It is a 2.4-trillion-parameter mixture-of-experts model accepting text, image, and video input and returning text, with a 1M-token context window, reasoning enabled by def Operational data shows Alibaba Cloud Int. as the sole host with $2/$6 per 1M tokens, $0.25 cache read, 39 tok/s P50 throughput, and 1.15s P50 latency. The page records a Sep 3, 2026 release date and an AutoExacto benchmark score of 63.6% attributed to Alibaba, while weighted-average effective input pricing lands at $0.

According to aireiter's developer-focused analysis, Qwen's September 2 announcement confirms that Qwen3.8-Max-0902 is a live revision with further post-training for coding and "cowork," exposed through the QwenCloud API. However, Alibaba Cloud's Model Studio page was updated the same day but still labels the model qwen The piece advises treating the -0902 suffix as a snapshot or revision of qwen3.8-max rather than a separately documented SKU, and warns that appending -0902 to existing requests should not be assumed to work. The documented envelope (2.4 trillion-parameter MoE, 1,000,000-token context, multimodal text output with funct

Command Code's directory entry names the exact model id qwen/qwen3.8-max-0902 and confirms the same $2 input, $6 output, $0.25 cache-read pricing with a 1M-token context window. It frames the release as an upgrade to Qwen 3.8 Max emphasizing stronger coding and agentic tool use, and surfaces blended-cost framing of rou The page explicitly marks intelligence, coding, and output-speed indices for the 0902 snapshot as not yet scored, indicating no independent third-party benchmark has been published. Comparison rows against Qwen 3.8 Max, Qwen 3.7 Max, DeepSeek V4 Pro, and DeepSeek V4 Flash give pricing, context, and (for non-0902 rows)

Alibaba released Qwen3.8-Max-0902 on September 2, 2026, as a post-training upgrade to Qwen3.8-Max focused on coding and long-horizon agent work. The update leaves the underlying architecture unchanged at 2.4 trillion parameters and a 1M token context, with pricing held flat at $2 per 1M input and $6 per 1M output. Data On the Code Arena WebDev leaderboard, Qwen3.8-Max-0902 takes the #1 spot with 1,691 points, narrowly ahead of Claude Opus 5 Max (1,687) and the prior Qwen3.8-Max (1,669). Eight coding benchmarks improve over the prior snapshot, with the largest jumps on TerminalBench 3.0 (11.3 to 29.0) and ProgramBench Almost Solved (1