Qwen 3.8 Max
LLM ModelsAlibaba's August 2026 flagship model, a 2.4T-parameter Mixture-of-Experts system with a 1M-token context window aimed at coding, agentic work, and multimodal tasks.
A giant specialist team with a million-token desk: it can draw on enormous capacity, but the useful question is which specialists are actually winning the tasks you care about.
Qwen 3.8 Max is Alibaba's August 2026 flagship large language model. It uses a 2.4 trillion parameter Mixture-of-Experts architecture and supports a one-million-token context window, positioning it for long-running coding, research, and agentic workflows.
Alibaba's launch chart compared Qwen 3.8 Max with Qwen 3.7 Max, Qwen 3.7 Plus, Claude Opus 4.8, Claude Fable 5, Gemini 3.1 Pro, and GPT-5.6 Sol across sixteen benchmarks. The model's strongest results came on vision, perception, and research-reproduction tasks including PaperBench, BabyVision, CharXiv, ERQA, PerceptionBench, LVBench, and OSWorld-Verified.
The same chart showed a more mixed picture on software engineering and long-horizon agent work. Claude Fable 5 led Qwen 3.8 Max on SWE-Pro, FrontierSWE, QwenReactBench, CoWorkBench, JobBench, Vision2Web, and MobileWorld, while GPT-5.6 Sol led on TerminalBench-2.1 and Agents' Last Exam. Alibaba notes that competing models were evaluated through their preferred harnesses, so cross-lab comparisons should be treated as directional rather than exact.
References & Resources
Related Terms
Last updated: August 3, 2026