Qwen Launches E-Commerce Bench to Test AI Agents in Long-Horizon Business Operations

Qwen Developers ·

Key Info

Qwen's E-Commerce Bench evaluates AI agents in long-horizon autonomous business operations: agents start with ¥100,000 and run an online store for 365 simulated days, handling sourcing, pricing, promotions, inventory and cash flow in a market driven by real e-commerce data.

Highlights

  • Built on real economic data: 6,886 products, 576 suppliers (152 of them fraudsters), daily storage fees, returns, and reputation mechanics.
  • Long-horizon reasoning is the core challenge: almost no model learns to buy cheaper or improve strategy over time, making it a useful reference for specialized business tasks.
Loading...