Alibaba Qwen Releases E-Commerce Bench: A 365-Day Stress Test for Autonomous Business Agents

Qwen Developers ·

Key Info

Alibaba Qwen unveiled E-Commerce Bench, a benchmark where AI agents run online stores for 365 days starting with ¥100,000, handling sourcing, negotiation, pricing, promotions, inventory and cash flow in a market built from real e-commerce data.

Highlights

  • Simulates real business conditions: 6,886 products, 576 suppliers (including 152 fraudsters), 600-minute workdays, storage fees, returns, and reputation effects.
  • Designed to test long-horizon autonomous operations, an area where most current models still struggle—early results suggest almost no model learns to buy cheaper over time.
  • Market dynamics are driven by real e-commerce data, making the benchmark more practical than synthetic business games.
Loading...