🔬 Tested by TokenRouter

Same prompt. Every model. Judge for yourself.

We run a fixed set of prompts against every model in the catalog and keep the outputs — so you can compare what they actually produce, not what the docs claim.

The prompt Remove the background from this image, leaving only the shapes on transparency.

gemini-2.5-flash-image output — gemini-2.5-flash-image google 7s

gemini-3-pro-image output — gemini-3-pro-image google 15s

gemini-3.1-flash-image output — gemini-3.1-flash-image google 5s

gemini-3.1-flash-lite-image output — gemini-3.1-flash-lite-image google ✗ failed 6s

OpenAI: GPT-5 Image output — OpenAI: GPT-5 Image openai 45s

OpenAI: GPT-5 Image Mini output — OpenAI: GPT-5 Image Mini openai 54s

OpenAI: GPT-5.4 Image 2 output — OpenAI: GPT-5.4 Image 2 openai 117s

Outputs generated 2026-07-17. Costs are measured from the provider's own billing where reported. Nothing here is simulated.