GPT-5.6 SOL tops e-commerce management study, Qwen team reports
Alibaba’s Qwen research team and the Hong Kong University of Science and Technology released results from E-Commerce Bench, a simulation testing AI models’ ability to run online businesses. The study, reported by ITmedia, gave 18 models 100,000 yuan each to operate virtual stores for 365 days. The simulation included 6,886 products, 576 suppliers, and 152 fraudulent firms designed to test fraud…
Key points
- GPT-5.6 SOL ranked first in E-Commerce Bench, growing 100,000 yuan to 1.43 million yuan.
- Anthropic’s Fable5 ranked second with 800,000 yuan, using fewer tool calls than GPT-5.6 SOL.
- Alibaba’s Qwen3.8 Max Preview was the top open-weight model, reaching 420,000 yuan.
OpenAI’s GPT-5.6 SOL ranked first, growing its initial funds to an average of 1.43 million yuan, a 14.3-fold increase. However, it ranked 16th in avoiding fraudulent suppliers. Anthropic’s Fable5 came second with final assets of about 800,000 yuan, achieving high profit efficiency with fewer tool calls. Among open-weight models, Alibaba’s Qwen3.8 Max Preview performed best, reaching about 420,000 yuan and demonstrating strong memory for past transaction prices.
The study found significant variation in AI capabilities, with 10 of 90 experiments ending in bankruptcy. Some models failed due to poor cash flow management, highlighting that while AI can run businesses over long periods, it still struggles with fraud detection and financial stability.
Model page: GPT-5.6 Sol →
The story so far
2 episodes →- GPT-5.6 SOL tops e-commerce management study, Qwen team reportsthis story
GPT-5.6 SOL tops AI online mall test after a year, study finds
digitaltoday.co.kr · 21 October 2026
Loading the full article…
This text was published by digitaltoday.co.kr and written by Choi Jae-won. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- PHP discloses critical PHP2Shell vulnerability affecting 70% of websites, discovered using GPT-5.6 Sol · 29 src
- Can AI spread copies of itself? Considering preparations based on Anthropic's 'self-replication' evaluation · 1 src
- UK AISI finds Mythos 5 kept targeting real people in July cyber eval despite evidence they were uninvolved · 3 src
- Harvard mathematicians debate whether OpenAI’s Navier-Stokes proof is understood · 5 src
- OpenAI collaborates with Poland's NASK on AI safety research · 5 src
Comments
via GitHub Discussions