Show HN: AdvertBench, ranking the ability of LLMs to create image ads

AdvertBench, a new benchmark ranking LLMs on their ability to create image ads, was released on Hacker News. The benchmark tests models by generating ads for a fictional company, Northstar Labs, and its product FlowPilot, evaluating visual style, composition, and text integration.

Which set of ads is better? Prompt Company: Northstar Labs. Product: FlowPilot, an AI productivity workspace for small teams. Audience: founders, operators, and project leads at teams of 5 to 50 people. Goal: drive trial signups by making the product feel practical, fast, and trustworthy. Required text: FlowPilot; Plan less. Ship more.; AI task planning for small teams; 14-day free trial. Logo: a clean four-point compass star inside a rounded square, paired with the FlowPilot wordmark. Visual style: premium SaaS ad with crisp dashboard UI elements, calm whitespace, sharp typography, subtle depth, deep navy, electric teal, white, and a warm yellow accent. Composition: show a planning dashboard or task board as the main visual, with the logo and offer text clearly readable.