Qwen Image 3 Pro takes OpenRouter's first benchmark

OpenRouter just published results from its first public benchmark — a meme-generation contest where image models went head-to-head on a single prompt. Qwen Image 3 Pro edged out the win with 35% of the vote. Nano Banana Pro, built by Google researchers, finished one vote back at 33%. The razor-thin margin left the outcome in the hands of the last voters.

The test used the "Why would I deceive you?" meme template — a character pointing at themselves with feigned innocence. Riverflow V2 Pro came third with 23%, while GPT Image 2 trailed far behind with only 9%.

OpenRouter aggregates access to dozens of models through a single API. OpenRouter ran the vote among its users. This is the first time the company has released results from a public meme-benchmark of this kind. Unlike standard benchmarks that measure technical accuracy or prompt adherence, the criterion was entirely subjective: which output was funniest or hit hardest.

The result shows how actual user preferences can contradict automated rankings or dry technical metrics. Qwen, from Alibaba researchers, keeps proving competitive in the open image-model arena. Nano Banana Pro showed it's still a serious contender despite the narrow loss. Next benchmark, anything goes.