Hi HN, I'm Shu, a software engineer and entrepreneur from Tokyo. With new image generation/editing models like Seedream 4.0, Nano Banana, and Qwen-Image releasing rapidly, I found it frustrating to evaluate them. Scattered APIs, inconsistent benchmarks, and pricey subscriptions (or complex setups) made side-by-side comparisons tough. I wanted a simple way to test Seedream 4.0 against Nano Banana without breaking the bank, so I built Model Arena over a few days.

Model Arena is a unified playground to experiment with SOTA image models. Upload an image and a prompt, and it runs Seedream 4.0, Nano Banana, and Qwen-Image in parallel, displaying outputs side-by-side for easy comparison. It’s built with Next.js, leveraging Firebase & Stripe for fast, private processing. No hefty subscriptions—just a credit base pricing to start exploring.

Unlike some existing tools that are expensive or limited to single-model testing, Model Arena focuses on accessibility and direct comparisons with clear metrics (e.g., generation speed, edit fidelity). It’s early days, but I’d love your feedback: What models do you want to see next? How do the outputs compare for your use cases? Try it out and let me know what you think!