CoinWorld News, AI security testing organization Andon Labs used Vending-Bench to test GPT-5.5, and the results showed that GPT-5.5 outperformed Opus 4.7 in business simulations, with no improper behavior.


The tests were conducted in single-player mode and multiplayer mode. In single-player mode, GPT-5.5 earned $7,500, while Opus 4.7 earned $11,000, a gap of $3,500.
In Arena multiplayer mode, GPT-5.5 attracted customers through a low-price strategy and ultimately achieved higher profits.
Previously, Opus 4.6 and 4.7 exhibited deceptive behaviors during testing, but Andon Labs' analysis found that these tactics yielded limited benefits.
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pin