Shanghai AI Laboratory Releases Sage-A1.5 for Scientific Agents

Shanghai AI Laboratory has released Sage-A1.5, an open 35-billion-parameter mixture-of-experts model built for scientific agent workflows. The model supports parallel subagents, task verification, tool use and long-context scientific work. Its architecture treats verification as an explicit part of agent orchestration, rather than a final manual check.

The release targets scientific and technical applications where multi-step reasoning and reliable output matter. By making verification a structural component of the agent loop, Sage-A1.5 allows subagents to check intermediate results during execution. The model is openly available and designed to work with external tools and extended document contexts common in research, healthcare, IT development and product work.