AI models struggle to manage virtual companies; Claude Fable 5 leads with $47M profit · 1 source tracked

A recent CEO-Bench competition, designed to test AI's ability to run a virtual SaaS startup, revealed mixed results. While many advanced AI models like GLM 5.1 and Gemini 3 Flash went bankrupt, Claude Fable 5 emerged as the top performer, generating $47.15 million. Notably, a purely rule-based algorithm also outperformed most LLMs, earning $15.76 million, suggesting that current AI models may struggle with the long-term strategic decision-making and uncertainty inherent in business management. AI

IMPACT Highlights the current limitations of AI in strategic decision-making and long-term planning, suggesting a need for specialized frameworks for different industries.

RANK_REASON Research paper detailing results of an AI competition simulating business management. [lever_c_demoted from research: ic=1 ai=1.0]

Read on 量子位 (QbitAI) →

• CEO-Bench
• Claude Fable 5
• Claude Opus 4.7
• Claude Opus 4.8
• DeepSeek V4 Pro
• Gemini 3 Flash
• GLM 5.1
• GPT-5.5
• Grok 4.20
• Kimi K2.6
• Qwen 3.7 Max

• product
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →