Alibaba’s Qwen team of AI researchers has unveiled Qwen3.8-Max, a new flagship 2.4-trillion-parameter mixture-of-experts multimodal large language model that targets autonomous software engineering and long-horizon enterprise work. The company states Qwen3.8-Max outperforms several leading proprietary models, including GPT-5.6 Sol Max and Fable 5, on some key benchmarks in agentic computing.
Qwen3.8-Max Benchmark Performance
Qwen3.8-Max scores 86.1 on the OSWorld-Verified benchmark, ahead of GPT-5.6 Sol Max and Fable 5. It also posts the highest reported score on PaperBench and leads or remains highly competitive across software engineering, research reproduction, multimodal reasoning, and visual web development benchmarks.
The model’s strongest differentiator is its ability to interact with operating systems, as measured by the OSWorld benchmark. This capability could automate business processes, including document processing, enterprise software integration, internal operations, and legacy workflows where APIs may not exist.
Related: GraphRAG outperforms vector method in some cases
Enterprise Automation and Research
Qwen3.8-Max combines many strengths into a single model aimed at enterprise automation. Alibaba positions the model as an autonomous coworker capable of executing projects that span days. Qwen3.8-Max can autonomously complete software projects lasting more than 10 days, reproduce research papers involving thousands of lines of code, perform iterative chip-design optimization, and continuously revise plans using multimodal feedback loops.
These demonstrations remain company-produced and have not yet been broadly replicated by independent evaluators. They illustrate a growing industry trend: frontier models are increasingly competing on their ability to finish entire workflows.
They are focusing on this aspect of the model.
Related: Cyber security must move at machine speed
Open Weights and Licensing
One important consideration for enterprises is the licensing terms that will govern Qwen3.8-Max’s open weights. The company says open weights for Qwen3.8-Max will be released next week, alongside Qwen3.8-27B. If that happens under a permissive license, it would represent the first time a Max-class Qwen model becomes available for self-hosted deployment—a move that could substantially reshape enterprise adoption. However, Alibaba has not yet disclosed the licensing terms, leaving open the possibility that the release could use a more restrictive custom license.
As the market for foundation models continues to evolve, it is likely that more models will prioritize autonomous execution and long-horizon planning. For organizations considering Qwen3.8-Max, the key will be to weigh its technical capabilities against the needs of their specific use cases and to carefully evaluate the licensing terms that will govern its open weights.
