The Sequence AI of the Week #725: Building Research, Not Answers: The DeepResearch Runtime
A deep dive into the newest model released by Alibaba's agentic group.
Alibaba’s Tongyi Lab is the research group behind a line of agent‑native systems—models, training curricula, and runtimes—built specifically for long‑horizon web research and tool‑use. Tongyi DeepResearch is its flagship open‑weights agentic model and stack. By contrast, Qwen is Alibaba Cloud’s broad, general‑purpose model family aimed at foundation capabilities—coding, math, multilingual tasks—delivered across multiple sizes and productized endpoints. In short: Tongyi Lab optimizes agent behavior and research workflows end‑to‑end (model + controller + iterative consolidation), while Qwen targets versatile base models that power many applications. This separation of concerns explains why DeepResearch ships with a full research runtime and artifact‑centric outputs, whereas Qwen ships as a family of foundation checkpoints and services.