Qwen 3.8 27B Outperforms Claude Opus 5 Medium on Agentic Index
The open-weights model Qwen 3.8 27B has outperformed Anthropic's proprietary Claude Opus 5 Medium on the Artificial Analysis Agentic Index. This benchmark evaluates large language models on agentic workflows, including planning, tool use, and autonomy. This result highlights the rapid advancement of open-source AI, showing that a relatively compact 27-billion-parameter model can match or exceed the capabilities of a massive, expensive proprietary model in complex agentic tasks. It could drive wider adoption of open-weights models for building autonomous AI agents. While Claude Opus 5 is Anthropic's flagship model designed for demanding reasoning and long-horizon agentic work, Qwen 3.8 27B achieved a higher score on the benchmark. The Agentic Index specifically measures behaviors like tool use, planning, autonomy, and complex problem solving.
## BACKGROUND
The Artificial Analysis Agentic Index is a composite benchmark that evaluates how well AI models function as agents, focusing on their ability to execute multi-step tasks independently. Claude Opus 5 is a state-of-the-art proprietary model developed by Anthropic, known for its high reasoning capabilities. Qwen is a series of open-weights large language models developed by Alibaba, which have consistently challenged proprietary models in various benchmarks.